CN111886619A

CN111886619A - Vehicle collision damage assessment method and system based on historical case

Info

Publication number: CN111886619A
Application number: CN201980021736.8A
Authority: CN
Inventors: 乐伟樑
Original assignee: Suzhou Shanshui Shuer Information Technology Co ltd
Current assignee: Suzhou Shanshui Shuer Information Technology Co ltd
Priority date: 2018-03-26
Filing date: 2019-03-26
Publication date: 2020-11-03
Also published as: WO2019184899A1

Abstract

The invention provides a damage assessment method after vehicle collision, which comprises the following steps: (a) obtaining a vehicle damage assessment historical case, and establishing a vehicle damage assessment historical case database; (b) receiving information of a vehicle to be damaged; (c) and calculating and determining the damage rating of the vehicle to be damaged according to the information of the vehicle to be damaged on the basis of the historical damage rating case data. The loss assessment method can rapidly, accurately and individually provide loss assessment amount by utilizing big data and a machine learning technology, and is not only suitable for independent use of each insurance company, but also suitable for a third-party loss assessment platform.

Description

Vehicle collision damage assessment method and system based on historical case

Technical Field

The invention relates to the field of vehicle insurance, in particular to a loss assessment method and system after vehicle collision.

Background

The process of estimating the loss after vehicle collision (called loss assessment for short) usually considers various technical factors, including the price of parts, the repair man-hour, the local labor rate, etc.

The damage assessment method relies on a combination of a damage assessment tool based on technical factors and subsequent macroscopic manual adjustment to achieve. Thus, in many cases, existing damage assessment methods and systems are used only as references, do not adapt to the dynamic changes of the market, and may result in price negotiations for multiple rounds. For example, using existing damage assessment systems sometimes takes days or even weeks to reach a consensus, and thus there are problems of inefficiency, high labor costs, and high error rates, reducing customer satisfaction.

Accordingly, there is a need for an improved impairment assessment method and system that overcomes one or more of the problems in the prior art described above.

Disclosure of Invention

One aspect of the invention provides a vehicle damage assessment method, adapted to be executed on a computer, the method comprising: (a) obtaining a vehicle damage assessment historical case, and establishing a vehicle damage assessment historical case database; (b) receiving information of a vehicle to be damaged; and (c) calculating the damage rating of the vehicle to be damaged according to the information of the vehicle to be damaged on the basis of the historical case data of damage rating.

In some embodiments, the method further comprises: and establishing a damage assessment model according to the vehicle damage assessment historical case. In the case of establishing the damage assessment model, the step (c) specifically includes: and calculating the damage rating of the vehicle to be damaged according to the information of the vehicle to be damaged and the damage rating model.

In some embodiments, the method further comprises: in the process of calculating the loss rating of the vehicle to be damaged, a self-evaluation confidence index for indicating the accuracy of the current loss rating is calculated.

In some embodiments, the method further comprises after step (c): and adding the information of the vehicle to be damaged and the damage rating of the vehicle to the historical case database of the vehicle damage rating.

In some embodiments, the method may further comprise: and after the information of the vehicles to be damaged and the damage rating thereof are added into the historical case database of vehicle damage rating, updating the damage rating model according to the information of the vehicles to be damaged and the damage rating thereof.

In some embodiments, the hierarchy of vehicle damage history case databases established by the method in step (a) includes a case level, a damage factor level, and a damage factor level.

In some embodiments, the vehicle damage history case database includes, but is not limited to, data associated with case levels and damage factors such as vehicle brand, vehicle series, vehicle type, year of vehicle production, damaged photograph of vehicle, damaged part name and number, damaged part picture, damaged part quantity, damaged part damage level, replacement part price, repair man-hour, local labor rate and damage rating. In some embodiments, the vehicle damage history case database further includes data associated with damage factor levels such as time of impact, location of impact, name of repair manufacturer, location of impact, and additional cost.

In some embodiments, the method further comprises: and preprocessing vehicle damage assessment historical case data in a vehicle damage assessment historical case database before building a damage assessment model according to the vehicle damage assessment historical case. The pre-treatment may be performed off-line. The preprocessing includes, but is not limited to, classifying, clustering, aggregating, sorting, summarizing, counting, cleaning, etc. the data, such as classifying the history cases according to the collision location.

In some embodiments, the establishing a damage assessment model according to the vehicle damage assessment historical case specifically includes: and (3) associating the damage assessment elements with the damage assessment factors through feature analysis and classification/clustering algorithms and/or machine learning to build a damage assessment model.

In some embodiments, the vehicle information to be damaged includes, but is not limited to, damaged vehicle information, damaged part information, including damage level, case affiliation information, repair information, and the like.

In some embodiments, the calculating the damage rating of the vehicle to be damaged according to the vehicle information to be damaged and the damage rating model includes: the calculation is carried out on a case level to obtain at least one loss assessment historical case with the similarity not lower than a preset similarity. In this case, the calculation results in at least one complete damage history case including reference payout amounts for a plurality of damage factors.

Alternatively, in some embodiments, the calculating the damage rating of the vehicle to be damaged according to the vehicle information to be damaged and the damage rating model includes: the calculation is carried out at the level of the damage assessment factor so as to obtain different historical data of one damage assessment factor of the vehicle to be damaged. In this case, the calculation obtains the reference payout amount for each damage factor of the vehicle to be damaged based on the historical payout amount in the damage factor.

Alternatively, in some embodiments, the calculating a damage rating of the vehicle to be damaged according to the vehicle information to be damaged and the damage rating model may include: and when the same or more than one damage history case and/or damage factor which is not lower than the preset similarity is/are obtained, sorting the damage history cases in time, and averaging the damage rating of the latest damage history case or cases and/or factor. In other embodiments, the damage rating is determined by other algorithms when more than one damage history case and/or factor is the same or not less than a predetermined similarity.

Alternatively, in other embodiments, calculating the damage rating of the vehicle to be damaged according to the vehicle information to be damaged and the damage rating model may include: the determination of the damage rating may involve the use of relevant data in other historical cases. For example, if historical claim data of one or more damaged parts corresponding to the vehicle to be damaged is absent in one or more historical case cases with high similarity, historical claim data of one or more same parts can be found from other historical cases (for example, historical cases with similarity lower than a predetermined similarity) and used for the damage rating of the part, and then the damage rating of the vehicle to be damaged is determined.

Alternatively, in some embodiments, in step (c), the calculation is performed at both the case level and the impairment factor level, i.e. the calculation does not involve the use of a single full historical case data. For example, the calculation determines the damage rating based on different categories/attributes of damage factor data from different historical cases.

In some embodiments, the calculating a confidence index indicating the accuracy of the current damage quota comprises: and according to the information (for example, one or more damage factors and damage factors can be included) of the vehicle to be damaged, searching for the factors such as the number and frequency of historical occurrences of the damage factors and the damage factors which are matched with the vehicle to be damaged, and the distribution and quality of historical settlement amount in the aggregate set of the damage factors, and calculating the confidence index.

In some embodiments, the method further comprises: determining whether the confidence index is below a predetermined threshold, and when the confidence index is below a threshold, generating alert data to trigger a manual process flow.

In some embodiments, the method further comprises: applying one or more adjustments based on the data of the at least one damage assessment historical case to determine a damage assessment for the vehicle to be damaged. In some embodiments, the one or more adjustments are applied based on one or more factors such as part price rate of change, manual rate of change, Key Performance Indicator (KPI) achievement rate, insurance service contribution rate, and the like. In some embodiments, the one or more adjustments comprise manual adjustments.

In some embodiments, the damage rating obtained by a conventional damage rating method can be used as a check point of the obtained damage rating to verify the accuracy of the method of the invention.

In some embodiments, the present invention provides a vehicle damage assessment system comprising a processor; and a memory for storing vehicle damage assessment instructions adapted to be loaded by the processor to perform any of the vehicle damage assessment methods described above.

In some embodiments, the present invention provides a computer readable non-transitory medium having computer readable instructions stored thereon, the instructions adapted to be loaded by a processor to perform any of the vehicle damage assessment methods described above.

According to the method provided by the invention, the rating of the vehicle to be damaged is calculated and determined on the basis of the historical case data and/or the factor data according to the information of the vehicle to be damaged and the record in the historical case database. The loss assessment method can rapidly, accurately and individually provide loss assessment amount by utilizing big data and a machine learning technology, and is not only suitable for independent use of each insurance company, but also suitable for a third-party loss assessment platform.

Drawings

FIG. 1 is a flow diagram of an exemplary method for damage assessment, according to one embodiment.

Fig. 2 is a flow diagram of an exemplary method of building a vehicle damage history case database, according to one embodiment.

FIG. 3 is a flow diagram of an exemplary method for damage assessment according to another embodiment.

Fig. 4 is a table showing an exemplary relationship of the damage assessment element to the damage assessment factor.

FIG. 5 is a block diagram of an exemplary impairment system, according to one embodiment.

FIG. 6 is a data structure diagram of a vehicle damage history case according to one embodiment.

FIG. 7 is a schematic diagram of a maintenance items table, according to one embodiment.

FIG. 8 is a schematic diagram of a damage assessment model according to one embodiment.

FIG. 9 is a schematic diagram of a machine learning algorithm, according to one embodiment.

FIG. 10 is a schematic loss assessment flow according to one embodiment.

Detailed Description

The invention will now be described in detail with reference to exemplary embodiments thereof, some of which are illustrated in the accompanying drawings. The following description refers to the accompanying drawings, in which like numerals refer to the same or similar elements throughout the different views unless otherwise specified. The aspects described in the following exemplary embodiments do not represent all aspects of the present invention. Rather, these aspects are merely examples of systems and methods according to various aspects of the present invention as recited in the appended claims.

Damage assessment based on technical factors needs to be determined according to a variety of technical factors after determining damaged parts and damage levels, which requires an insurance company or a third party to continuously maintain a database of various vehicle brands, years, models, part types, part prices, maintenance man-hours, labor rates, corresponding maintenance logics, and the like. However, there are challenges to achieving a damage rating that is acceptable to both insurance companies and repair partners. For example, insurance companies and damage assessment tool providers often have difficulty obtaining an accurate and high coverage database as the basis for damage assessment, third party sources of data often are not approved, and so on. These large volumes of data (e.g., various vehicle make, year, model, part type, part price, man-hours of maintenance, labor rates, etc.), wide coverage, quick updates, and difficulty in reaching a consensus (e.g., on man-hours of maintenance) are challenges for accurate damage assessment. Furthermore, in china, for example, insurance companies actually still adopt a macroscopic claim management mode, rather than a case-based microscopic precise damage assessment mode, which considers many non-technical factors. For example, if a dealer servicing a vehicle to be damaged (e.g., a 4S store) may bring up a large amount of vehicle insurance business, the insurance company may modify the parameters accordingly or manually adjust the damage results. As another example, if the insurance company's annual KPI (Key Performance indicator) has been met, the damage may relax up. This macro-management model of insurance companies may result in neglecting individual case-level fine-grained management.

FIG. 1 shows an exemplary method 100 for vehicle damage assessment according to one embodiment of the present invention. The method 100 begins at step 110 by building a vehicle damage history case database. The implementation of the exemplary method 100 is based on the building of a vehicle damage history case database. The insurance companies form a large amount of historical case data in the long-term vehicle damage claim business, and the damage rating of the historical cases is calculated by a conventional technology-factor-based damage rating tool and is combined with manual adjustment based on non-technology factors according to a conventional damage rating mode. Therefore, the damage rating of the historical cases is the truest reflection of the market situation, the experience and knowledge of the damage raters and the nuclear loss raters are precipitated, the damage rating of the vehicle to be damaged is determined based on the damage rating of the historical cases, and the accuracy of the damage rating can be ensured.

The method of building the historical case database may be implemented using database building techniques common in the art, for example, a relational database may be used. Method 200 illustratively shows a method of building a historical case database. The method 200 begins by defining a damage factor 210 for a historical case of vehicle damage. The damage factor may include one or more of a brand of vehicle, a series of vehicles, a type of vehicle, a collision location, a part name, a number of parts, a price of parts, a repair man-hour, a local labor rate, a time of collision occurrence, a collision occurrence location, a repair maker name, an additional expense, a vehicle year of production, a version of a type of vehicle, a color of a body, a degree of damage to a part, and a body decor. The impairment factors described above may be represented using a data structure used by a relational database. In some specific embodiments, the impairment factors may be represented in the form of one or more database tables. In step 220, the damage assessment elements of the vehicle damage assessment history case are defined. In some embodiments, the damage assessment elements can include crash types including, but not limited to, right front case, right rear case, left front case, right front case, left side case, right side case, left rear case, right rear case, mild case, severe case, or various combination cases thereof. In other embodiments, the damage component may include, but is not limited to, damage vehicle information (brand, model, year, configuration, etc.), damaged part information (parts, degree of damage, etc.), affiliated insurance agency information (region, discount rate, etc.), repair information (repair shop level, parts source, discount rate, etc.), and the like.

Steps

210 and 220 may also be performed in the reverse order. Finally, in step 230, the one or more impairment components are associated with one or more impairment factors, respectively. For example, one impairment factor is associated with a plurality of impairment components. As another example, each of the impairment components share one or more of the same impairment factors. The relationship between the impairment factors and the impairment components is exemplarily shown in the table 400 shown in fig. 4.

These impairment factors, as well as impairment components, may be updated, e.g., added, deleted, or modified, so that the database can be maintained from more, updated dimensions. When the method of the present invention is implemented on a third party damage assessment platform, historical cases from multiple insurance companies may be aggregated and a historical case database built according to the exemplary method described above.

In a specific embodiment, as shown in fig. 6, the damage assessment component may include: vehicle model/configuration, damaged parts, degree of damage, source of parts, parts discount rate, and man-hour discount rate, etc. The impairment factors may include: the loss items may include loss items, maintenance items, disassembly and assembly items, and painting items, wherein the loss items may include a loss item name (e.g., loss item 1), a part number, a unit price, a usage amount, and a part fee. The associated attribute information included in the loss item is stored in the relational database in an associated manner, and the attribute information may be corresponding attribute values, for example, the value of the configuration number may be an identification number of a combination of letters and numbers, and the values of the unit price, the usage amount, and the component fee may be arabic numbers. The maintenance items may include a maintenance item name (e.g., maintenance item 1), a work type, a maintenance level, a maintenance man-hour, and a maintenance fee. The attribute information associated with the maintenance items is stored in the database in an associated manner, for example, the names of the attributes can be used as keys, and the values corresponding to the attributes are used as values.

In some embodiments, the vehicle damage history case database may include a claims information table, a branch office table, a repair shop information table, a damaged vehicle information table, and a maintenance items table. For example, the claims information table may include in the database the information shown in simplified table 1, including: loss order number, time of occurrence, place of occurrence, organization, repair shop, final plan time, VIN code, vehicle type, and cause of occurrence.

TABLE 1

The branch table may include information in the database as shown in simplified table 2, including: an organization, a 4S store discount rate, a market price discount rate, and an applicable price discount rate.

TABLE 2

Mechanism	Discount rate for 4S store	Market price discount rate	Discount rate of applicable price
Suzhou division Co Ltd	85.00	65.00	80.00

The repair shop information table may include information in the database as shown in simplified table 3, including: a repair shop code, a repair shop name, an association mechanism, a repair shop type, an accessory channel, an accessory discount rate, a man-hour unit price type, and the like.

TABLE 3

The damaged vehicle information table may include information in the database shown in simplified table 4, including: number plate type, engine number, new vehicle purchase price, date of first class, actual value, vehicle use property, country, long plate, vehicle series, vehicle type, style, power source, engine type, transmission type, displacement, overturn, whether the air bag can run or not, loss degree, main collision point, secondary collision point and the like.

TABLE 4

Table 4 (continuation)

Table 4 (continuation)

The maintenance items table may include information in the database as shown in fig. 7, including: repair shop, loss assessment project name, accessory number, operation type, reference unit price, dosage, folded accessory, work type, maintenance degree, folded maintenance, reference disassembly and assembly, paint spraying type, reference paint spraying, accessory cost, accessory external repair, accessory residual value, accessory depreciation, management rate, disassembly and assembly superposition discount rate, paint spraying superposition discount rate and the like.

The historical cases in the vehicle damage assessment historical database provided by the embodiment of the application can be divided from different layers, for example, the historical cases can be claim information, damage assessment elements and damage assessment factors. For example, the above partitioning of historical cases may be accomplished using machine learning. For example, as shown in fig. 6, from the claims themselves, the historical case data may include accident vehicle base information, repair shop information, and insurance branch information, wherein the accident vehicle base information may include: VIN code, engine number, vehicle type, purchase price of new vehicle, initial registration date of driving license, vehicle use property, occurrence reason, accident handling mode, accident type, collision grade, loss photo and the like, the insurance branch office can set different accessories and time discount according to different locations, and the repair shop information can comprise repair shop code, repair shop name, repair shop type, associated mechanism, cooperation type and the like. From the damage assessment factors, the historical case data may include: vehicle/configuration, damaged parts, degree of damage, source of parts, parts discount rate, formula discount rate, and the like. From the damage factor, the historical case data may include: the change item, the auxiliary material item, the sheet metal item, the dismounting item, the painting item and the like of each claim and the claim payment amount of each claim, wherein each item can be associated with the damage assessment element in a classified mode. The damage factor is a damage factor that determines and affects the amount of a payout of the damage factor. The loss assessment factors in a large number of historical claims are associated with the loss assessment factors, and then the loss assessment pricing reference value of the loss assessment factors under the condition of the same/similar loss assessment factors can be calculated, so that a loss assessment model is obtained. And calculating the damage rating of the vehicle to be damaged according to the information of the vehicle to be damaged and the damage rating model.

In some embodiments, the damage factors in the historical cases are classified according to the damage factors, and a set of damage factors for the same damage factor can be calculated, wherein one or more historical payout amounts in the set reflect historical payout for the same damage factor. And calculating and processing the set of the claims amount by using rules and algorithms to obtain the reference claims amount of the damage factor in the set of the damage factors. Reference claim calculation rules and algorithms include, but are not limited to, median; taking an average value; when the number of identical payments in the collection exceeds a predetermined proportion (e.g. 50%), taking the number; removing singular values, such as the first 10% and the last 10% of a damage-assessment factor payment in descending order; and so on. Based on the table damage rating model, a damage rating of the vehicle to be damaged may be calculated. FIG. 8 is a simplified view of the damage-assessment model, wherein the left columns 2-4 (name of the damage-assessment project, number of parts, and operation type) are historical case damage-assessment factors, the columns 5-8 (branch company, type of repair shop, channel of parts, unit price type of man-hour) are damage-assessment elements, the columns 9-11 (part fee, assembly fee, and paint fee) are the amount of the historical damage-assessment factors paid, and the columns 12-14 are the calculated reference amount of the damage-assessment factors.

In some embodiments, the impairment model may be established by machine learning, using algorithms including, but not limited to, neural networks, reinforcement learning neural networks, generation networks in antagonistic neural networks, and the like. The first stage is a model generation stage, and the specific structure in the neural network can be set by a person skilled in the art according to the actual requirements of the damage assessment factors, such as parts, maintenance, painting, disassembly and assembly, part contexts, maintenance unit contexts, accessory channels, regions, branches, and the like, which is not limited in the embodiment of the present invention. In the embodiment of the invention, the historical payout amount of the loss assessment factors and the corresponding loss assessment factors is input into the machine learning algorithm for learning and training, and the reference price (including parts and working hours) of each item in the corresponding loss assessment scheme is output to establish a loss assessment model. The second stage is a discriminant model stage in which the rationality of the price predicted in the first stage is judged based on a deep learning model, and the outputs of the stages are 0 and 1, i.e., whether the price is rational or not. In the training process, the aim of generating the model is to generate each price in the loss assessment plan as much as possible to deceive the discriminant model. The aim of the discriminant model is to make the price obtained by generating the model and the real price respectively. Thus, the generation model and the discrimination model form a dynamic 'game process'. Finally, a price sufficient for reality is generated.

In a specific embodiment, the machine learning algorithm model may be as shown in fig. 9. The machine learning algorithm combines a multi-layer deep learning model for generating an antagonistic network and a neural network to perform the price prediction of the claim unit. As shown in fig. 9, the left side is a generated model in which a part set and a maintenance item set of a claim form are constructed, and all parts and maintenance items appearing in the claim form construct an N-dimensional input vector containing N parts and maintenance items, where N is an integer greater than 1; and deep learning is carried out through the Encoder to finally generate a predicted price corresponding to the N-dimensional input vector, meanwhile, real claim list (GI) data is mixed and sent into a right-side countermeasure network model, and the generated predicted price is judged through a deep learning model of the Decoder part, so that the rationality of the generated price is judged, and a confidence index is obtained.

Due to the integrity of the historical cases and the deviation in the actual damage rating, the confidence index for indicating the accuracy of the damage rating at this time is calculated by the method, and comprises the following steps: and according to the information (for example, one or more damage factors and damage factors can be included) of the vehicle to be damaged, searching the factors such as the number and frequency of historical occurrences of the damage factors and damage factors of the vehicle to be damaged and the distribution and quality of historical settlement amount in the damage factor aggregation set, and calculating the confidence index. For example, in FIG. 7, when a damage factor in a damage claim does not appear in the damage model, the confidence index of the damage factor is 0; the historical payout amount variance of the factor is inversely related to the confidence index, and so on. In a historical case, when the pay amount corresponding to a certain loss assessment factor is in a relatively fixed value interval, the distribution of the loss assessment factor is considered to be relatively stable, and accordingly, the confidence index is improved; in the historical case, the confidence index in the case where a certain damage factor occurs only twice and the amount of the two payments differs greatly is also low. And combining the confidence indexes of each loss assessment factor in the loss assessment claim according to the proportion of the paid amount to obtain the loss assessment confidence indexes of the loss assessment claim.

In some embodiments, in order to solve the problem that the damage assessment factors in the damage assessment claims cannot match with proper damage reference values in the calculation process when the number of historical claims is insufficient, a clustering algorithm is adopted to cluster similar and similar variables of certain damage assessment factors, for example, in the process of calculating labor hours, branch organizations or regions with the same or similar labor hour standards are clustered, so that the success rate and the accuracy rate of damage assessment of the damage assessment factors are improved.

In some embodiments, in order to solve the problem that the damage factors in the damage claims cannot match with proper damage reference values in the calculation process when the number of historical claims is insufficient, a conditional search method is adopted, for example, a set of the damage factors of the most ideal damage factors is firstly searched, for example, the search fails, a suboptimal set of the calculated set is searched, for example, the set fails again, and the set is searched again until the success or the complete failure, so that the success rate and the accuracy rate of damage factor damage evaluation are improved. For each level of successful calculation, a corresponding confidence index is assigned, such as 1 for one-step success and 0 for complete failure. The method can improve the success rate and the accuracy rate of loss assessment.

In some embodiments, it is determined whether the confidence index is below a predetermined threshold, and when the confidence index is below a threshold, alert data is generated to trigger a manual process flow.

In some embodiments, the result of artificial damage assessment is added into the vehicle damage assessment historical case database, and the damage assessment model is updated according to the information of the vehicle to be damaged and the damage assessment amount thereof. For example, the artifact will add to the damage assessment model and give higher computational weight to adjust the reference payout for the damage assessment factor.

In step 120, the method receives information about the vehicle to be damaged. Such relevant information may include one or more damage factors and damage factors, such as license plate number, vehicle brand, vehicle series, vehicle type, collision location, damaged part name and number, vehicle year of production, version of vehicle type, body color, degree of damage to parts, and body decor. As one skilled in the art can appreciate, the more sufficient the information about vehicles to be damaged is provided, the easier it is to match the records in the historical case database, so that a more matching claim/damage record is retrieved, providing a basis for outputting cases with high similarity. Of course, on the other hand, matching of cases also depends on the breadth and completeness of the records in the database. In actual practice, the input of relevant information for the vehicle to be damaged may not rely on the entry of data on a case-by-case basis. For example, in some cases, when entering the license plate number of a vehicle to be damaged, the method 100 may automatically retrieve information from an official platform or other platform about the corresponding brand of the vehicle, the train, the model, the year the vehicle was produced, the version of the model, the color of the body, the decoration of the body, etc.

In step 130, the vehicle to be damaged is computed from the historical case database to find similar historical cases. The calculation comprises traversing the vehicle damage assessment historical case database, and calculating the similarity between the vehicle to be damaged and each case or/and factor in the vehicle damage assessment historical case database according to a preset calculation mode and based on the received information of the vehicle to be damaged. The predetermined calculation mode may be a machine learning algorithm, or may be determined as needed by those skilled in the art. One exemplary mode of computation is to compute a weighted sum of the similarity of the values of the impairment factors. Another exemplary mode of computation is to compute a weighted average of the similarity of the values of the impairment factors. In the exemplary calculation mode, the person skilled in the art needs to determine the weight of each damage factor, for example, the weight of the brand, the series, and the model of the vehicle may be higher than the weight of the year of the vehicle production, the version of the model, the color of the body, or the weight of the damage degree of the collision part and the part may be higher than the weight of the collision place. The determination of the weights can be determined by a limited number of tests by a person skilled in the art.

In some cases, the resulting historical cases are ranked from high to low in similarity. Alternatively, only history cases above a predetermined similarity threshold are output with the predetermined similarity as the threshold. And when more than one damage history case is not lower than the preset similarity, sorting the damage history cases in time. The predetermined similarity can be set and adjusted by a person skilled in the art according to the actual situation. In some cases, the similarity can be adjusted in real time, and the output similar cases are correspondingly increased or decreased in real time, so that the user can determine the number of similar historical cases according to actual cases.

In step 140, a damage rating is calculated based on the obtained data of the historical cases. An exemplary algorithm determines the damage rating of the vehicle to be damaged based on the obtained damage rating or an average thereof for one or more damage rating history cases. When there are multiple similar damage history cases above the similarity threshold, the nearest case or cases may be selected to determine the damage rating (e.g., average) of the vehicles to be damaged. Alternatively, when there are a plurality of similar damage-rating history cases above the threshold, the damage rating of one of the history cases may be used as the standard. For example, if a certain historical case matches exactly the information of the vehicle to be damaged, it may not be necessary to average as described above.

Another exemplary method of determining the damage rating of a vehicle to be damaged involves the use of relevant data in other historical cases. For example, if historical claim data of one or more damaged parts corresponding to the vehicle to be damaged is absent in one or more high-similarity historical cases, historical claim data of one or more same parts can be found from other historical cases and used for determining the damage rating of the part, and then the damage rating of the vehicle to be damaged is determined.

Another exemplary method of determining the damage rating of a vehicle to be damaged does not involve the use of a single full historical case data. In this case, the method 100 determines the damage rating based on data from different categories/attributes of different historical cases.

In step 140, one or more adjustments may be applied over the determined damage rating to ultimately determine the damage rating of the vehicle to be damaged. These adjustments are applied, for example, based on one or more factors such as part price rate of change, labor rate of change, KPI achievement rate, insurance business contribution rate, and the like. These adjustments may be computer-implemented. For example, the adjustment is made according to the rate of change of the part price, for example, according to the following formula: the average change rate of the price of the accessories in the previous period of the reference price of the accessories and the historical claims.

The adjustment is carried out according to the rate of change of the manual rate, and can be carried out according to the following formula: and the labor cost change rate of the average preposed time period of the historical claims is the reference price of the labor cost.

Because the historical loss rating case is a case before a certain time, the historical loss rating case has a certain hysteresis in time, and the loss rating made by referring to the historical case has a certain error, the historical loss rating case is adjusted by using the current factors such as the change rate of the part price, the change rate of the artificial rate, the KPI standard-reaching rate, the insurance service contribution rate and the like on the loss rating made by referring to the historical case, and the accuracy and the real-time performance of determining the loss rating can be improved.

Method 300 illustrates another exemplary impairment method. The method 300 begins by building a vehicle damage history case database (step 310), which is substantially identical to step 110, but during the building process, the method 300 further pre-processes the database. The pretreatment can be performed generally off-line/off-line. The data preprocessing can be based on a big data analysis and aggregation algorithm and can also be carried out according to different purposes. For example, history cases can be classified based on the type of collision, so when matching the history case database, matching records can be screened out more quickly and accurately according to the type of collision without traversing all data records. Steps 320 through 340 are substantially the same as steps 120 through 140 of method 100. In step 350, the method 300 adds the damage-rated vehicle information and its damage-rated amount to the historical case database, enriching and updating the records of the database to provide more accurate and timely matching records for the damage-rating of new cases. When the data update is performed on the historical case database (step 352), the data preprocessing step therein needs to be performed again. Thus, the data pre-processing is performed dynamically in nature. When the present invention is implemented in the form of artificial intelligence and machine learning, the addition of new data records may cause the machine to dynamically adjust its algorithms, thereby continuously improving accuracy.

A specific embodiment of the present disclosure is described below in conjunction with fig. 10. Rectangular box 1000 identifies the architecture of the damage assessment system provided by the present disclosure. And (3) constructing a vehicle damage assessment historical case according to the massive damage assessment lists which are already finished (step 1001), and then performing big data machine learning damage assessment model according to the damage assessment historical case (step 1002). After the damage model learning, repair inventory data for the vehicle to be damaged is received, which may include one or more repair items, for example. These repair inventory data are acquired by image recognition of images of the manual survey and/or collision site. And (6) carrying out artificial intelligence algorithm damage assessment according to the damage assessment module obtained by machine learning and the maintenance list data of the current damage assessment vehicle (step 1003), and obtaining a subentry damage assessment list and an intelligent confidence index (step 1004). If the confidence index is greater than a predetermined threshold, then automatic damage assessment may be performed (step 1005), otherwise, human intervention damage assessment may be performed (step 1006). The results of the artificial damage assessment can be put into damage assessment model learning to update the damage assessment model (step 1007). After automatic damage assessment, the results of the damage assessment are put into historical case learning.

Fig. 5 shows a block diagram of a damage assessment system 500 according to some disclosed embodiments. The system may include a processor 521, an input/output (I/O) device 522, a memory 523, a storage 526, a database 527, and a display 528.

The processor 521 may be one or more known processing devices, such as those manufactured by Intel^TMProduced Pentium^TMSeries of microprocessors, or from AMD^TMManufactured Turion^TMA series of microprocessors. Processor 521 may include a single core processor system or a multi-core processor system capable of parallel processing. For example, processor 521 may be a single-core processor with virtual processing techniques. In some embodiments, processor 521 may utilize a logical processor to execute and control multiple processes simultaneously. The processor 521 may execute virtual machine technology, or other similar known technologies, to enable execution, control, enable, manipulate, store, etc. of a plurality of software processes, applications, programs, etc. In another embodiment, processor 521 includes a multi-core processor configuration (e.g., dual or quad core) configured to provide parallel processing functionality, allowing for a lossy system500 execute multiple processes simultaneously. Those skilled in the art will appreciate that other types of processor configurations may be implemented to provide the functionality described herein.

Memory 523 may include one or more storage devices configured to store instructions used by processor 521 to perform the functions of the disclosed embodiments. For example, memory 523 may be configured with one or more software instructions, such as instructions 524, which when executed by processor 521, may perform one or more operations. The disclosed embodiments are not limited to a single program or computer configured to perform specialized tasks. For example, the memory 523 may include a single instruction 524 that performs the functions of the impairment system 500, or the instruction 524 may include multiple instructions.

Memory 523 may also store data 525, which data 525 may reflect any type of information in any form that performs the functions in the disclosed embodiments. For example, data 525 may include metadata of a historical case database related to the damage assessment calculations, as well as other data that enables processor 521 to perform the functions in the disclosed embodiments.

The I/O device 522 may be configured to allow data to be received and/or transmitted. The I/O devices 522 may include one or more digital and/or analog communication devices that allow the impairment system 500 to communicate with other machines and devices. The damage assessment system 500 can also include one or more databases 527, or be communicatively coupled to one or more databases 527 via a network. For example, database 527 may comprise Oracle^TMDatabase, Sybase^TMA database, or other relational or non-relational database, such as a Hadoop sequence file, HBase, or Cassandra. In an exemplary embodiment, database 527 may store historical case data for damage fixes. This metadata may be created by a user, for example, and stored in database 527.

The present invention also provides a computer readable medium having stored thereon computer readable instructions adapted to be loaded by a processor to perform any of the vehicle damage assessment methods described herein. The computer-readable medium may include a removable medium as a package medium including a magnetic disk (including a flexible disk), an optical disk (including a CD-ROM (compact disc-read only memory) and a DVD (digital versatile disc)), a magneto-optical disk (including an MD (mini disc)), or a semiconductor memory. In some embodiments, the computer readable medium resides, for example, in an application store to provide an application, such as a mobile terminal application, that encodes any of the vehicle damage assessment methods illustrated in the present invention.

Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. The scope of the invention is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the invention and including such departures from the present disclosure as come within known or customary practice in the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.

It will be appreciated that the invention is not limited to the precise arrangements described above and illustrated in the drawings and that various modifications and changes can be made without departing from the scope of the invention. The scope of the invention is only limited by the claims.

Claims

A vehicle damage assessment method adapted to be executed on a computer, the method comprising:

(a) obtaining a vehicle damage assessment historical case, and establishing a vehicle damage assessment historical case database;

(b) receiving information of a vehicle to be damaged; and

(c) and calculating and determining the damage rating of the vehicle to be damaged according to the information of the vehicle to be damaged on the basis of the historical damage rating case data.
The method of claim 1, wherein the method further comprises after step (c): (d) and adding the information of the vehicle to be damaged and the damage rating of the vehicle to the historical case database of the vehicle damage rating.
The method of claim 1, wherein the data in the vehicle damage history case database is preprocessed in step (a).
The method of claim 1 wherein the hierarchy of vehicle damage history case databases established in step (a) includes a case level, a damage factor level, and a damage factor level.
The method of claim 4, wherein in step (c), the calculating is performed at a case level, resulting in at least one loss-rated historical case not less than a predetermined similarity.
The method of claim 4, wherein in step (c) the calculation is performed at a impairment factor level to obtain different historical data for at least one impairment factor.
The method of claim 4, wherein in step (c) the calculation is performed at both a case level and a impairment contributor level.
The method of claim 1, involving applying one or more adjustments in step (c) to determine a final damage rating for the vehicle to be damaged.
The method of claim 8, wherein the one or more adjustments include but are imposed according to one or more factors such as part price rate of change, manual rate of change, KPI achievement rate, insurance business contribution rate, or the one or more adjustments include manual adjustments.
The method of claim 1, further comprising:

after step (a), building a damage assessment model according to the vehicle damage assessment historical case.
The method of claim 10, the method comprising:

and calculating the damage rating of the vehicle to be damaged according to the information of the vehicle to be damaged and the damage rating model.
The method of claim 10, further comprising:

and after the information of the vehicles to be damaged and the damage rating thereof are added into the historical case database of vehicle damage rating, updating the damage rating model according to the information of the vehicles to be damaged and the damage rating thereof.
The method of claim 10, further comprising:

and (3) associating the damage assessment elements with the damage assessment factors through feature analysis and classification/clustering algorithms and/or machine learning to build a damage assessment model.
The method of claim 10, further comprising:

and in the process of calculating the loss rating of the vehicle to be damaged, calculating a confidence index for indicating the accuracy of the current loss rating.
The method of claim 14, said calculating a confidence index indicating accuracy of the present damage rating comprising:

and searching the historical occurrence times and frequency of the damage factors and the damage factors of the vehicle to be damaged and the distribution and quality of the historical settlement amount in the damage factor aggregation set according to the information of the vehicle to be damaged, and calculating the confidence index.
A vehicle damage assessment system, comprising:

a processor; and

a memory for storing vehicle damage assessment instructions adapted to be loaded by a processor to perform the method of any of claims 1 to 15.
A computer-readable non-transitory medium storing computer-readable instructions adapted to be loaded by a processor to perform the method of any of claims 1 to 15.