An intelligent car insurance pricing system based on multi-modal data processing

Abstract

The application relates to an intelligent car insurance quotation system based on multi-modal data processing, which comprises a multi-source data acquisition module, a risk domain construction module and a dynamic quotation module; the multi-source data acquisition module analyzes driving behavior characteristics, vehicle condition characteristics and driving environment characteristics of a target vehicle based on collected image data, structured text data and sensor time series data; the risk domain construction module constructs a driving behavior risk curve of the target vehicle based on the driving behavior characteristics, constructs a vehicle condition risk curve of the target vehicle based on the vehicle condition characteristics, then constructs a driving environment risk curve of the target vehicle based on the driving environment characteristics, and finally constructs a dynamic quotation domain of the target vehicle based on the driving behavior risk curve, the vehicle condition risk curve and the driving environment risk curve; and the dynamic quotation module performs quotation analysis on the dynamic quotation domain to obtain a car insurance quotation of the target vehicle.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data and car insurance pricing, and particularly relates to an intelligent car insurance pricing system based on multi-modal data processing. BACKGROUND

[0002] With the rapid development of the insurance financial field, especially in the current high-speed development of the Internet economy era, the traditional vehicle insurance underwriting system cannot meet the expansion of actual insurance business. At present, the traditional car insurance method is that the business personnel expand the business, telephone sales and car owners buy insurance in the car, all of which cannot be separated from the car insurance pricing, which has the following shortcomings:

[0003] The traditional car insurance pricing relies on artificial experience and single data source (such as vehicle model, historical accident record), which has the following problems: single data dimension: it is difficult to capture the actual state of the vehicle (accident damage) or driving behavior risk; subjective bias: artificial evaluation is low in efficiency and is easily affected by subjective factors; dynamic update lag: unable to real-time integrate the latest vehicle state or driving data. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides an intelligent car insurance pricing system based on multi-modal data processing, which comprises a multi-source data acquisition module, a risk domain construction module and a dynamic pricing module, and the modules are communicatively connected; the risk domain construction module comprises a first unit, a second unit, a third unit and a construction unit, and the units are communicatively connected;

[0005] The multi-source data acquisition module acquires image data, structured text data and sensor time series data of the target vehicle, and analyzes driving behavior characteristics, vehicle condition characteristics and driving environment characteristics of the target vehicle based on the image data, structured text data and sensor time series data;

[0006] The first unit of the risk domain construction module constructs a driving behavior risk curve of the target vehicle based on the driving behavior characteristics;

[0007] The second unit of the risk domain construction module constructs a vehicle condition risk curve of the target vehicle based on the vehicle condition characteristics;

[0008] The third unit of the risk domain construction module constructs a driving environment risk curve of the target vehicle based on the driving environment characteristics;

[0009] The construction unit of the risk domain construction module constructs a dynamic pricing domain of the target vehicle based on the driving behavior risk curve, the vehicle condition risk curve and the driving environment risk curve;

[0010] The dynamic pricing module performs pricing analysis on the dynamic pricing domain to obtain the car insurance pricing of the target vehicle.

[0011] According to a preferred embodiment, the image data comprises: vehicle damage images, vehicle body high-definition detail images, and driving environment images collected in real time by a vehicle-mounted camera; the structured text data comprises vehicle repair records, insurance policy history, and fault diagnosis reports; and the sensor time series data comprises vehicle state parameters collected at a high frequency through a vehicle-mounted OBD interface, GPS positioning trajectories, and environmental sensor data.

[0012] According to a preferred embodiment, the analysis based on the image data, structured text data, and sensor time series data to obtain driving behavior characteristics, vehicle condition characteristics, and driving environment characteristics of the target vehicle comprises:

[0013] Data cleaning, dimension unification, and missing value filling are performed on the image data, structured text data, and sensor time series data to obtain standard image data, standard text data, and standard time series data;

[0014] Visual feature analysis, text semantic mining, and behavior pattern extraction are respectively performed on the standard image data, standard text data, and standard time series data to obtain image features, text features, and time series features;

[0015] Temporal and spatial alignment layers and attention fusion mechanisms are used on the image features, text features, and time series features to obtain a spatiotemporally synchronized fusion feature tensor;

[0016] The spatiotemporally synchronized fusion feature tensor is used to perform vehicle state diagnosis, driving risk prediction, and environment perception verification on the target vehicle to obtain the driving behavior characteristics, vehicle condition characteristics, and driving environment characteristics of the target vehicle.

[0017] According to a preferred embodiment, the construction of a driving behavior risk curve of the target vehicle based on the driving behavior characteristics comprises:

[0018] A plurality of target driving behavior feature points are obtained based on the driving behavior characteristics of the target vehicle; standard driving behavior features are obtained from a database, and a plurality of corresponding standard driving behavior feature points are obtained based on the standard driving behavior features;

[0019] The number of accidents and the amount of accidents of the target vehicle are obtained based on historical accident records of the target vehicle, and a compensation correction distance is calculated using a preset compensation correction function and the number of accidents and the amount of accidents of the target vehicle;

[0020] A first risk buffer domain is constructed with each standard driving behavior feature point as the center and the compensation correction distance as the radius, and a circle tangent to the first risk buffer domain is found with each target driving behavior feature point as the center, and then the corresponding tangent point is taken as the risk critical point corresponding to each target driving behavior feature point;

[0021] connecting all the risk critical points corresponding to the target driving behavior feature points to obtain a driving behavior risk curve.

[0022] According to a preferred embodiment, constructing a vehicle condition risk curve of the target vehicle based on the vehicle condition features of the target vehicle comprises:

[0023] obtaining a plurality of target vehicle condition feature points based on the vehicle condition features of the target vehicle; obtaining a plurality of standard vehicle condition feature points based on the standard vehicle condition features obtained from the database;

[0024] obtaining the number of accidents and the accident amount of the target vehicle based on the historical accident records of the target vehicle, and calculating a compensation correction distance using a preset compensation correction function and the number of accidents and the accident amount of the target vehicle;

[0025] constructing a second risk buffer domain with each standard vehicle condition feature point as the center and the compensation correction distance as the radius, and finding a circle tangent to the second risk buffer domain with each target vehicle condition feature point as the center, and then taking the corresponding tangent point as the risk critical point corresponding to each target vehicle condition feature point;

[0026] connecting all the risk critical points corresponding to the target vehicle condition feature points to obtain a vehicle condition risk curve.

[0027] According to a preferred embodiment, constructing a driving environment risk curve of the target vehicle based on the driving environment features of the target vehicle comprises:

[0028] obtaining a plurality of target driving environment feature points based on the driving environment features of the target vehicle; obtaining a plurality of standard driving environment feature points based on the standard driving environment features obtained from the database;

[0029] obtaining the number of accidents and the accident amount of the target vehicle based on the historical accident records of the target vehicle, and calculating a compensation correction distance using a preset compensation correction function and the number of accidents and the accident amount of the target vehicle;

[0030] constructing a third risk buffer domain with each standard driving environment feature point as the center and the compensation correction distance as the radius, and finding a circle tangent to the third risk buffer domain with each target driving environment feature point as the center, and then taking the corresponding tangent point as the risk critical point corresponding to each target driving environment feature point;

[0031] connecting all the risk critical points corresponding to the target driving environment feature points to obtain a driving environment risk curve.

[0032] According to a preferred embodiment, constructing the dynamic offer domain of the target vehicle based on the driving behavior risk curve, the vehicle condition risk curve and the driving environment risk curve comprises:

[0033] Stepwise boundary convergence is performed on the driving behavior risk curve to obtain a driving behavior risk boundary line;

[0034] Stepwise boundary convergence is performed on the vehicle condition risk curve to obtain a vehicle condition risk boundary line;

[0035] Stepwise boundary convergence is performed on the driving environment risk curve to obtain a driving environment risk boundary line;

[0036] Dynamic fitting is performed on the driving behavior risk boundary line, the vehicle condition risk boundary line and the driving environment risk boundary line to obtain the dynamic offer domain.

[0037] According to a preferred embodiment, the stepwise boundary convergence performed on the driving behavior risk curve to obtain the driving behavior risk boundary line comprises:

[0038] All risk critical points of the driving behavior risk curve are obtained, and a feature value of each risk critical point is extracted; an average feature value of all risk critical points is calculated, and a first convergence reference point is obtained based on the average feature value;

[0039] All extreme points of the driving behavior risk curve are obtained, and all extreme points are taken as first convergence points to obtain a plurality of first convergence points;

[0040] A boundary iteration direction of each first convergence point is taken as a direction of each first convergence point pointing to the first convergence reference point;

[0041] The first convergence point is moved in the boundary iteration direction of each first convergence point by a preset step length, and it is judged whether the first convergence point is in a region of a first risk buffer domain; if the first convergence point is in the region of the first risk buffer domain, the first convergence point is continuously moved by the preset step length until the first convergence point is not in the region of the first risk buffer domain, and the first convergence point at this time is taken as a driving behavior risk boundary point;

[0042] The driving behavior risk boundary points of all first convergence points are connected to obtain the driving behavior risk boundary line.

[0043] According to a preferred embodiment, the stepwise boundary convergence performed on the vehicle condition risk curve to obtain the vehicle condition risk boundary line comprises:

[0044] All risk critical points of the vehicle condition risk curve are obtained, and a feature value of each risk critical point is extracted; an average feature value of all risk critical points is calculated, and a second convergence reference point is obtained based on the average feature value;

[0045] all extreme points of the vehicle condition risk curve as second convergence points to obtain a plurality of second convergence points;

[0046] a direction of each second convergence point pointing to the second convergence reference point as a boundary iteration direction of each second convergence point;

[0047] moving the second convergence point in the boundary iteration direction of each second convergence point according to a preset step size, and determining whether the second convergence point is within a region of the second risk buffer domain; if the second convergence point is within the region of the second risk buffer domain, continuing to move the second convergence point according to the preset step size until the second convergence point is not within the region of the second risk buffer domain, and taking the second convergence point at this time as a vehicle condition risk boundary point;

[0048] connecting the vehicle condition risk boundary points of all second convergence points to obtain a vehicle condition risk boundary line.

[0049] According to a preferred embodiment, the step-by-step boundary convergence of the driving environment risk curve to obtain a driving environment risk boundary line comprises:

[0050] obtaining all risk critical points of the driving environment risk curve, and extracting a feature value of each risk critical point; calculating an average feature value of all risk critical points, and obtaining a third convergence reference point based on the average feature value;

[0051] obtaining all extreme points of the driving environment risk curve, and taking all extreme points as third convergence points to obtain a plurality of third convergence points;

[0052] a direction of each third convergence point pointing to the third convergence reference point as a boundary iteration direction of each third convergence point;

[0053] moving the third convergence point in the boundary iteration direction of each third convergence point according to a preset step size, and determining whether the third convergence point is within a region of the third risk buffer domain; if the third convergence point is within the region of the third risk buffer domain, continuing to move the third convergence point according to the preset step size until the third convergence point is not within the region of the third risk buffer domain, and taking the third convergence point at this time as a driving environment risk boundary point;

[0054] connecting the driving environment risk boundary points of all third convergence points to obtain a driving environment risk boundary line.

[0055] According to a preferred embodiment, the intelligent vehicle insurance pricing system further comprises a user terminal; the user terminal sends a pricing request to the multi-source data acquisition module; the multi-source data acquisition module determines a target vehicle in response to the received pricing request, and acquires image data, structured text data and sensor time series data of the target vehicle.

[0056] According to a preferred embodiment, the dynamic pricing module performs pricing analysis on the dynamic pricing field to obtain the vehicle insurance pricing of the target vehicle, and sends the vehicle insurance pricing of the target vehicle to the corresponding user terminal.

[0057] According to a preferred embodiment, the device with communication and data transmission functions for the terminal to use for the user includes a smartphone, a tablet computer, a notebook computer and a smart watch.

[0058] The application has the following beneficial effects: the application breaks through the limitation of traditional vehicle insurance relying on a single static index by deeply fusing multi-dimensional data such as driving behavior, vehicle state and driving environment, and constructs a dynamic risk assessment model. Based on real-time interaction, the synergistic influence of driving habit risk tendency, vehicle component health degree and environmental road condition complexity is analyzed to realize accurate premium calculation and instant adjustment, and significantly improve risk identification coverage and response efficiency. At the same time, by cross-verification of the matching of vehicle use intensity and mechanical loss, the abnormal behavior monitoring capability is strengthened, and the fraud risk is effectively prevented. The technology promotes the transformation of vehicle insurance pricing from an empirical extensive mode to a personalized dynamic quantification, provides a protection scheme suitable for the actual risk of each vehicle, forms a whole-chain active risk management closed loop of “evaluation-warning-optimization”, and has the advantages of the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 A structural block diagram of an intelligent vehicle insurance pricing system based on multi-modal data processing is provided for an exemplary embodiment. DETAILED DESCRIPTION

[0060] The exemplary embodiments will be described in detail herein with reference to the attached drawings. In the following description, the same numbers are used to designate the same elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not representative of all embodiments consistent with the present application. Rather, they are merely examples of devices and methods consistent with some aspects of the present application, as detailed in the appended claims.

[0061] The terms used in the present application are merely for the purpose of describing specific embodiments, and are not intended to limit the present application. The singular forms “a”, “an” and “the” used in the present application and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” used herein means and includes any or all possible combinations of one or more associated listed items.

[0062] It should be understood that, although the terms first, second, third, etc. can be employed in describing various information used or presented herein, these information are not to be construed as being limited to these terms. Such terms can be only used to, for example, distinguish one information from another. For example, a first information could later be referred to as a second information without departing from the scope of the present application. As used herein, the term "if' can be construed to mean "when" or "upon" or "in response to determining" terms in contexts of examples. It is to be understood that the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise.

[0063] Referring to Figure 1 In one embodiment, an intelligent car insurance pricing system based on multi-modal data processing includes a multi-source data acquisition module, a risk domain construction module, and a dynamic pricing module, which are communicatively connected; the risk domain construction module includes a first unit, a second unit, a third unit, and a construction unit, which are communicatively connected.

[0064] The multi-source data acquisition module acquires image data, structured text data, and sensor time series data of the target vehicle, and analyzes driving behavior characteristics, vehicle condition characteristics, and driving environment characteristics of the target vehicle based on the image data, structured text data, and sensor time series data;

[0065] The image data includes vehicle damage images, high-definition detail images of the vehicle body, and driving environment images collected in real time by the vehicle-mounted camera; the structured text data includes vehicle repair records, insurance policy history, and fault diagnosis reports; and the sensor time series data includes vehicle state parameters collected at a high frequency through the vehicle-mounted OBD interface, GPS positioning trajectory, and environmental sensor data.

[0066] The image data is sourced from a vehicle-mounted camera (resolution ≥ 1920x1080) and user-uploaded accident photos (format: JPEG or PNG, size ≤ 10MB);

[0067] The text data is sourced from insurance company databases (JSON / CSV format) and OCR-recognized repair records;

[0068] The sensor time series data is collected through the OBD interface at a frequency of 10 Hz, covering sudden acceleration (acceleration > 2.5 m / s 2 ), sudden braking (deceleration > 3.0 m / s 2 ) events

[0069] The structured text data includes vehicle repair records (including repair items, dates, and costs), insurance policy history (insurance date, claim amount), and fault diagnosis reports (fault code and description), which are stored in JSON / CSV format and conform to the ISO18542 standardized machine-readable data structure.

[0070] The sensor time series data includes vehicle state parameters (such as engine speed, vehicle speed, sudden acceleration / sudden braking events) collected at a high frequency (10 Hz) through the vehicle OBD interface, GPS positioning trajectory (accuracy ≤1 meter), and environmental sensor data (such as tire pressure, temperature), and is stored in a standardized time series format for driving behavior modeling and risk analysis.

[0071] In a preferred embodiment, the driving behavior characteristics, vehicle condition characteristics, and driving environment characteristics of the target vehicle obtained based on the image data, structured text data, and sensor time series data analysis include:

[0072] The image data, structured text data, and sensor time series data are subjected to data cleaning, dimension unification, and missing value filling to obtain standard image data, standard text data, and standard time series data.

[0073] The standard image data, standard text data, and standard time series data are respectively subjected to visual feature analysis, text semantic mining, and behavior pattern extraction to obtain image features, text features, and time series features.

[0074] The image features, text features, and time series features are subjected to spatio-temporal alignment layer and attention fusion mechanism to obtain a spatio-temporal synchronous fusion feature tensor.

[0075] The spatio-temporal synchronous fusion feature tensor is used for vehicle state diagnosis, driving risk prediction, and environment perception verification of the target vehicle to obtain the driving behavior characteristics, vehicle condition characteristics, and driving environment characteristics of the target vehicle.

[0076] Optionally, the driving behavior characteristics are used to represent the driving habits of the driver of the target vehicle, and the driving habits include dangerous driving behaviors such as frequent sudden braking, speeding, random lane changing, following too close, night driving, and fatigue driving.

[0077] Optionally, the vehicle condition characteristics are used to represent the vehicle condition of the target vehicle, and the vehicle condition covers appearance scratches, engine oil leakage or abnormal noise, transmission hesitation, brake sensitivity, tire wear, chassis rust or deformation, and whether the electronic devices (such as ABS and airbags) are normal, fully reflecting the vehicle health and potential risks.

[0078] Optionally, the driving environment characteristics are used to represent the environment in which the target vehicle frequently drives, including the weather (rain / snow / fog) of the frequently driven area, the road type (highway / mountain road / non-paved road), the congestion frequency, the night driving proportion, and whether it passes through a school, a construction zone, or a sharp and steep road section, affecting the vehicle wear and accident probability.

[0079] The first unit of the risk domain construction module constructs a driving behavior risk curve of the target vehicle based on the driving behavior characteristics.

[0080] In a preferred embodiment, constructing the driving behavior risk curve of the target vehicle based on the driving behavior features comprises:

[0081] Obtaining a plurality of target driving behavior feature points based on the driving behavior features of the target vehicle; obtaining a plurality of standard driving behavior feature points based on the standard driving behavior features obtained from the database;

[0082] Obtaining the number of accidents and the amount of accidents of the target vehicle based on the historical accident records of the target vehicle, and calculating a compensation correction distance using a preset compensation correction function and the number of accidents and the amount of accidents of the target vehicle;

[0083] Constructing a first risk buffer domain with each standard driving behavior feature point as the center and the compensation correction distance as the radius, and finding a circle tangent to the first risk buffer domain with each target driving behavior feature point as the center, and then taking the corresponding tangent point as the risk critical point corresponding to each target driving behavior feature point;

[0084] Connecting all the risk critical points corresponding to the target driving behavior feature points to obtain the driving behavior risk curve.

[0085] Optionally, the standard driving behavior features are related features of standard driving behaviors, and the standard driving behaviors refer to the standard driving habits of strictly abiding by traffic regulations, maintaining reasonable vehicle distance, gentle acceleration and deceleration, avoiding overspeed or dangerous lane changing, and actively adjusting vehicle speed and operation according to weather and road conditions.

[0086] The second unit of the risk domain construction module constructs a vehicle condition risk curve of the target vehicle based on the vehicle condition features.

[0087] In a preferred embodiment, constructing the vehicle condition risk curve of the target vehicle based on the vehicle condition features comprises:

[0088] Obtaining a plurality of target vehicle condition feature points based on the vehicle condition features of the target vehicle; obtaining a plurality of standard vehicle condition feature points based on the standard vehicle condition features obtained from the database;

[0089] Obtaining the number of accidents and the amount of accidents of the target vehicle based on the historical accident records of the target vehicle, and calculating a compensation correction distance using a preset compensation correction function and the number of accidents and the amount of accidents of the target vehicle;

[0090] Constructing a second risk buffer domain with each standard vehicle condition feature point as the center and the compensation correction distance as the radius, and finding a circle tangent to the second risk buffer domain with each target vehicle condition feature point as the center, and then taking the corresponding tangent point as the risk critical point corresponding to each target vehicle condition feature point;

[0091] connecting all the risk critical points corresponding to the target vehicle condition feature points to obtain a vehicle condition risk curve.

[0092] The standard vehicle condition feature is a feature of a standard vehicle condition, which refers to a perfect state of a vehicle appearance without damage, mechanical parts (such as an engine, a brake system, and a tire) without wear or aging, and all functions (light, electronic equipment) meeting the safety standards of the factory.

[0093] The third unit of the risk domain construction module constructs a driving environment risk curve of the target vehicle based on the driving environment features.

[0094] In a preferred embodiment, constructing the driving environment risk curve of the target vehicle based on the driving environment features comprises:

[0095] obtaining a plurality of target driving environment feature points based on the driving environment features of the target vehicle; obtaining a plurality of standard driving environment feature points based on standard driving environment features from a database;

[0096] obtaining the number of accidents and the amount of accidents of the target vehicle based on the historical accident records of the target vehicle, and calculating a compensation correction distance using a preset compensation correction function and the number of accidents and the amount of accidents of the target vehicle;

[0097] constructing a third risk buffer domain with each standard driving environment feature point as the center and the compensation correction distance as the radius, and finding a circle tangent to the third risk buffer domain with each target driving environment feature point as the center, and then taking the corresponding tangent point as the risk critical point corresponding to each target driving environment feature point;

[0098] connecting all the risk critical points corresponding to the target driving environment feature points to obtain a driving environment risk curve.

[0099] Optionally, the standard driving environment feature is a feature related to a standard driving environment, which refers to a low-risk driving environment with a smooth and unobstructed road, good weather conditions (no extreme rain / snow / fog), avoidance of construction zones, sharp curves and steep slopes, moderate traffic flow, and meeting the basic conditions for safe driving of the vehicle.

[0100] Optionally, the compensation correction distance is used to correct the number of accidents and the amount of accidents of the user when evaluating the driving behavior risk, the vehicle state risk, and the driving environment risk of the target vehicle. For example, although the driving behavior and the driving environment of the driver of the target vehicle are not up to standard, the driver has never been in an accident for many years, so the insurance premium cannot be increased solely based on the substandard driving behavior and driving environment.

[0101] The constructing unit of the risk domain constructing module constructs a dynamic pricing domain of the target vehicle based on the driving behavior risk curve, the vehicle condition risk curve and the driving environment risk curve.

[0102] Optionally, the constructing of the dynamic pricing domain of the target vehicle based on the driving behavior risk curve, the vehicle condition risk curve and the driving environment risk curve comprises:

[0103] Stepwise boundary convergence is performed on the driving behavior risk curve to obtain a driving behavior risk boundary line;

[0104] Stepwise boundary convergence is performed on the vehicle condition risk curve to obtain a vehicle condition risk boundary line;

[0105] Stepwise boundary convergence is performed on the driving environment risk curve to obtain a driving environment risk boundary line;

[0106] Dynamic fitting is performed on the driving behavior risk boundary line, the vehicle condition risk boundary line and the driving environment risk boundary line to obtain the dynamic pricing domain.

[0107] In a preferred embodiment, the stepwise boundary convergence performed on the driving behavior risk curve to obtain the driving behavior risk boundary line comprises:

[0108] All risk critical points of the driving behavior risk curve are obtained, and feature values of each risk critical point are extracted; average feature values of all risk critical points are calculated, and a first convergence reference point is obtained based on the average feature values;

[0109] All extreme points of the driving behavior risk curve are obtained, and all the extreme points are taken as first convergence points to obtain a plurality of first convergence points;

[0110] A boundary iteration direction of each first convergence point is taken as a direction of each first convergence point pointing to the first convergence reference point;

[0111] The first convergence point is moved in the boundary iteration direction of each first convergence point by a preset step length, and it is judged whether the first convergence point is in a region of a first risk buffer domain; if the first convergence point is in the region of the first risk buffer domain, the first convergence point is continuously moved by the preset step length until the first convergence point is not in the region of the first risk buffer domain, and the first convergence point at this time is taken as a driving behavior risk boundary point;

[0112] All the driving behavior risk boundary points of all the first convergence points are connected to obtain the driving behavior risk boundary line.

[0113] In a preferred embodiment, the stepwise boundary convergence performed on the vehicle condition risk curve to obtain the vehicle condition risk boundary line comprises:

[0114] all extreme points of the vehicle condition risk curve are obtained as the second convergence points to obtain a plurality of second convergence points;

[0115] all extreme points of the vehicle condition risk curve are obtained as the second convergence points to obtain a plurality of second convergence points;

[0116] a direction of each second convergence point pointing to the second convergence reference point is taken as a boundary iteration direction of each second convergence point;

[0117] the second convergence point is moved in the boundary iteration direction of each second convergence point according to a preset step size, and it is judged whether the second convergence point is in the region of the second risk buffer domain; if the second convergence point is in the region of the second risk buffer domain, the second convergence point is continuously moved according to the preset step size until the second convergence point is not in the region of the second risk buffer domain, and the second convergence point at this time is taken as a vehicle condition risk boundary point;

[0118] all vehicle condition risk boundary points of all second convergence points are connected to obtain a vehicle condition risk boundary line.

[0119] In a preferred embodiment, the driving environment risk curve is subjected to step-by-step boundary convergence to obtain a driving environment risk boundary line, which comprises:

[0120] all extreme points of the vehicle condition risk curve are obtained as the second convergence points to obtain a plurality of second convergence points;

[0121] all extreme points of the vehicle condition risk curve are obtained as the second convergence points to obtain a plurality of second convergence points;

[0122] a direction of each second convergence point pointing to the second convergence reference point is taken as a boundary iteration direction of each second convergence point;

[0123] the second convergence point is moved in the boundary iteration direction of each second convergence point according to a preset step size, and it is judged whether the second convergence point is in the region of the second risk buffer domain; if the second convergence point is in the region of the second risk buffer domain, the second convergence point is continuously moved according to the preset step size until the second convergence point is not in the region of the second risk buffer domain, and the second convergence point at this time is taken as a vehicle condition risk boundary point;

[0124] all vehicle condition risk boundary points of all second convergence points are connected to obtain a vehicle condition risk boundary line.

[0125] Optionally, the extreme points are maximum values or minimum values in a certain region of the curve.

[0126] Optionally, each feature point has a unique corresponding feature value, and the corresponding feature point can be uniquely located by the feature value.

[0127] The dynamic pricing module performs pricing analysis on the dynamic pricing domain to obtain the vehicle insurance pricing of the target vehicle.

[0128] Optionally, the vehicle insurance pricing of the target vehicle includes the cost of the vehicle insurance that the target vehicle needs to pay in the current year.

[0129] Preferably, the dynamic pricing domain is used to represent the driving behavior risk distribution, vehicle condition risk distribution and driving environment risk distribution of the target vehicle.

[0130] Preferably, the intelligent vehicle insurance pricing system further comprises a user terminal; the user terminal sends a pricing request to the multi-source data acquisition module; the multi-source data acquisition module determines the target vehicle in response to the received pricing request, and acquires image data, structured text data and sensor time series data of the target vehicle.

[0131] Preferably, the dynamic pricing module performs pricing analysis on the dynamic pricing domain to obtain the vehicle insurance pricing of the target vehicle, and sends the vehicle insurance pricing of the target vehicle to the corresponding user terminal.

[0132] Preferably, the terminal used by the user is a device with communication function and data transmission function, including a smartphone, a tablet computer, a notebook computer and a smart watch.

[0133] The present application breaks through the limitation of traditional vehicle insurance relying on single static index by deeply integrating multi-dimensional data such as driving behavior, vehicle state and driving environment, and constructs a dynamic risk assessment model. Based on real-time interaction, the synergistic influence of driving habit risk tendency, vehicle component health degree and environmental road condition complexity is analyzed to realize accurate premium calculation and instant adjustment, significantly improving risk identification coverage and response efficiency. At the same time, through cross verification of the matching of vehicle use intensity and mechanical loss, the abnormal behavior monitoring capability is strengthened, and the fraud risk is effectively prevented. This technology promotes the transformation of vehicle insurance pricing from an empirical extensive mode to a personalized dynamic quantification, provides a protection scheme adapted to the actual risk of each vehicle, and forms a full-chain active risk management closed loop of "evaluation-warning-optimization".

[0134] Computer readable program instructions for carrying out operations of the present application can be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present application.

[0135] The foregoing is considered as illustrative only of the principles of the application. Further, since numerous modifications and changes will readily occur to those skilled in the art, it is not desired to limit the application to the exact construction and practice described. Accordingly, all such variations are intended to be included within the scope of present application as defined in the following claims.

Claims

1. An intelligent auto insurance quotation system based on multimodal data processing, characterized in that, The intelligent auto insurance pricing system includes: a multi-source data acquisition module, a risk domain construction module, and a dynamic pricing module, with communication connections between the modules; the risk domain construction module includes a first unit, a second unit, a third unit, and a construction unit, with communication connections between the units; The multi-source data acquisition module collects image data, structured text data, and sensor time-series data of the target vehicle, and analyzes the driving behavior characteristics, vehicle condition characteristics, and driving environment characteristics of the target vehicle based on the image data, structured text data, and sensor time-series data. The first unit of the risk domain construction module constructs a driving behavior risk curve for the target vehicle based on the driving behavior characteristics. The second unit of the risk domain construction module constructs a vehicle condition risk curve for the target vehicle based on the vehicle condition characteristics. The third unit of the risk domain construction module constructs the driving environment risk curve of the target vehicle based on the driving environment characteristics; The construction unit of the risk domain construction module constructs the dynamic pricing domain of the target vehicle based on the driving behavior risk curve, vehicle condition risk curve, and driving environment risk curve; The dynamic pricing module performs pricing analysis on the dynamic pricing domain to obtain the car insurance quote for the target vehicle; The driving behavior risk curve of the target vehicle constructed based on the aforementioned driving behavior characteristics includes: Several target driving behavior feature points are obtained based on the driving behavior characteristics of the target vehicle; standard driving behavior features are obtained from the database, and several corresponding standard driving behavior feature points are obtained based on the standard driving behavior features; Based on the historical accident records of the target vehicle, the number of accidents and the amount of accidents of the target vehicle are obtained, and the compensation correction distance is calculated using a preset compensation correction function and the number of accidents and the amount of accidents of the target vehicle. A first risk buffer zone is constructed with each standard driving behavior feature point as the center and the compensation correction distance as the radius. A circle tangent to the first risk buffer zone is found with each target driving behavior feature point as the center. The corresponding tangent point is then used as the risk critical point for each target driving behavior feature point. Connect the risk thresholds corresponding to all target driving behavior feature points to obtain the driving behavior risk curve; A dynamic pricing domain for the target vehicle is constructed based on the driving behavior risk curve, vehicle condition risk curve, and driving environment risk curve; including: The driving behavior risk boundary line is obtained by step-by-step boundary convergence of the driving behavior risk curve; The vehicle condition risk boundary line is obtained by step-by-step boundary convergence of the vehicle condition risk curve. The driving environment risk boundary line is obtained by step-by-step boundary convergence of the driving environment risk curve; The dynamic pricing domain is obtained by dynamically fitting the risk boundary lines of driving behavior, vehicle condition, and driving environment. The driving behavior risk boundary line is obtained by performing step-by-step boundary convergence on the driving behavior risk curve, including: Obtain all risk thresholds of the driving behavior risk curve and extract the feature value of each risk threshold; calculate the average feature value of all risk thresholds and obtain the first convergence benchmark point based on the average feature value; Obtain all extreme points of the driving behavior risk curve, and use all extreme points as the first convergence points to obtain several first convergence points; The direction from each first convergence point to the first convergence reference point is taken as the boundary iteration direction of each first convergence point; Move the first convergence point along the boundary iteration direction of each first convergence point by a preset step size, and determine whether the first convergence point is within the area of ​​the first risk buffer zone. If the first convergence point is within the area of ​​the first risk buffer zone, continue to move the first convergence point by a preset step size until the first convergence point is no longer within the area of ​​the first risk buffer zone, and take the first convergence point at this time as the driving behavior risk boundary point. Connect all the driving behavior risk boundary points of the first convergence point to obtain the driving behavior risk boundary line.

2. The intelligent auto insurance quotation system according to claim 1, characterized in that, The image data includes: vehicle damage images, high-definition detailed images of the vehicle body, and driving environment images collected in real time by the vehicle-mounted camera; the structured text data includes vehicle maintenance records, insurance policy history, and fault diagnosis reports; the sensor time-series data includes vehicle status parameters, GPS positioning trajectory, and environmental sensor data collected at high frequency through the vehicle-mounted OBD interface.

3. The intelligent auto insurance quotation system according to claim 1, characterized in that, Based on the analysis of the image data, structured text data, and sensor time-series data, the driving behavior characteristics, vehicle condition characteristics, and driving environment characteristics of the target vehicle are obtained, including: Data cleaning, dimensional unification, and missing value imputation are performed on image data, structured text data, and sensor time series data to obtain standard image data, standard text data, and standard time series data; Visual feature analysis, text semantic mining, and behavioral pattern extraction were performed on standard image data, standard text data, and standard time-series data respectively to obtain image features, text features, and time-series features; Spatiotemporal alignment layers and attention fusion mechanisms are applied to image features, text features, and temporal features to obtain a spatiotemporally synchronized fused feature tensor; By utilizing spatiotemporally synchronized fusion feature tensors, vehicle status diagnosis, driving risk prediction, and environmental perception verification are performed on the target vehicle to obtain the target vehicle's driving behavior characteristics, vehicle status characteristics, and driving environment characteristics.

4. The intelligent auto insurance quotation system according to claim 1, characterized in that, The intelligent auto insurance quotation system also includes a user terminal; the user terminal sends a quotation request to the multi-source data acquisition module; the multi-source data acquisition module responds to the received quotation request to determine the target vehicle and collects the target vehicle's image data, structured text data and sensor time-series data.

5. The intelligent auto insurance quotation system according to claim 4, characterized in that, The dynamic pricing module analyzes the dynamic pricing domain to obtain the car insurance quote for the target vehicle and sends the car insurance quote for the target vehicle to the corresponding user terminal.

6. The intelligent auto insurance quotation system according to claim 5, characterized in that, The user terminal is a device used by a user that has communication and data transmission functions, including: smartphones, tablets, laptops, and smartwatches.

Images