Accident vehicle part loss assessment price determination method and related equipment

Abstract

The invention discloses a method and device for determining the loss assessment price of a spare part of an accident vehicle, electronic equipment, a computer readable storage medium and a computer program product, and aims to solve the problem that the loss assessment price determination method provided by the prior art may cause that the loss assessment price difference of the same spare part in different loss assessment places is too large and is unreasonable. The method comprises the steps of obtaining target loss assessment price band information corresponding to information of a target part by querying a pre-established corresponding relationship between information of different parts and corresponding loss assessment price band information according to the information of the target part of which the loss assessment price is to be determined; the loss assessment price band information in the corresponding relation is determined according to a plurality of localized price representation vectors with similarity; representing historical loss assessment prices of the same part in different loss assessment places corresponding to different localized feature representation vectors of the same part; based on the information influencing the loss assessment discount rate of the target loss assessment land, predicting the loss assessment discount rate of the target part in the target loss assessment land; and determining the loss assessment price of the target part in the target loss assessment land according to the target loss assessment price band information and the predicted loss assessment discount rate.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, electronic device, computer-readable storage medium and computer program product for determining the damage assessment price of spare parts for an accident vehicle. Background Art

[0002] After an accident occurs and the insured vehicle is damaged, the insurance company needs to make a claim for the insured vehicle. Most of the cost of car insurance claims comes from the damage assessment price of the insured vehicle's replacement parts. The process of determining the damage assessment price is as follows: Collect the price data of various spare parts provided by multiple manufacturers of spare parts; manually or automatically review and clean the price data - for example, identify and remove outliers, duplicates and data entries that do not meet the specifications in the price data; take the provinces in China as examples, and adjust the reviewed and cleaned price data locally - including considering the consumption level, tax policy, logistics cost and other factors of each damage assessment location, and make appropriate corrections and calibrations to the price data to ensure that the corrected and calibrated price data can truly reflect the market conditions of each damage assessment location; for each damage assessment location, encapsulate and store the corresponding localized adjusted price data in accordance with specific formats and standards to form a localized price benchmark data set for each damage assessment location; in the subsequent actual damage assessment process, set different damage assessment discount rates according to the degree of cooperation between the insurance company and the 4S stores (or repair shops) in each damage assessment location - the damage assessment discount rate is used for damage assessment and can be called the damage assessment discount rate. The price data of the spare parts in the localized price benchmark data set of each damage assessment location is multiplied by the corresponding damage assessment discount rate to obtain the damage assessment price of the spare parts in each damage assessment location. The main defects of the above-mentioned price assessment method are: due to the influence of consumption levels, tax policies or logistics costs in different damage assessment places (such as provinces), the damage assessment prices of the same spare parts in different damage assessment places may be too different and unreasonable. Summary of the invention

[0003] The embodiment of the present application provides a method for determining the damage appraisal price of accident vehicle spare parts, so as to solve the problem that the damage appraisal price determination method provided by the prior art may result in unreasonable large differences in the damage appraisal prices of the same spare parts in different damage appraisal locations.

[0004] The embodiments of the present application also provide a device for determining the damage assessment price of spare parts for an accident vehicle, an electronic device, a computer-readable storage medium, and a computer program product.

[0005] The present application embodiment adopts the following technical solutions: A method for determining the damage assessment price of spare parts of an accident vehicle, the method comprising: According to the information of the target spare parts for which the damage assessment price is to be determined, the target damage assessment price band information corresponding to the information of the target spare parts is obtained by querying the correspondence relationship between the information of different spare parts and the corresponding damage assessment price band information established in advance; the damage assessment price band information in the correspondence relationship is determined based on a number of localized price representation vectors that are similar to each other; different localized feature representation vectors corresponding to the same spare part represent the historical damage assessment prices of the same spare part in different damage assessment locations; Based on information affecting the damage assessment discount rate of the target damage assessment location, predicting the damage assessment discount rate of the target spare part at the target damage assessment location; The damage assessment price of the target spare part at the target damage assessment location is determined based on the target damage assessment price band information and the predicted damage assessment discount rate.

[0006] A device for determining the damage assessment price of spare parts of an accident vehicle, the device comprising: The price band determination unit is used to obtain the target damage assessment price band information corresponding to the information of the target spare part by querying the pre-established correspondence relationship between the information of different spare parts and the corresponding damage assessment price band information according to the information of the target spare part for which the damage assessment price is to be determined; the damage assessment price band information in the correspondence relationship is determined according to a number of localized price characterization vectors having similarities with each other; different localized feature characterization vectors corresponding to the same spare part characterize the historical damage assessment prices of the same spare part in different damage assessment locations; A discount rate prediction unit, configured to predict the damage discount rate of the target spare part at the target damage assessment location based on information affecting the damage discount rate at the target damage assessment location; The damage assessment price determination unit is used to determine the damage assessment price of the target spare part at the target damage assessment location based on the target damage assessment price band information and the predicted damage assessment discount rate.

[0007] An electronic device, comprising: a memory and a processor, wherein: The memory is used to store programs; the processor is coupled to the memory and is used to execute the programs stored in the memory so as to execute the above-mentioned method for determining the damage assessment price of spare parts for accident vehicles.

[0008] A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer, can implement the above-mentioned method for determining the damage assessment price of spare parts for an accident vehicle.

[0009] A computer program product stores instructions, which, when executed by a computer, enable the computer to implement the above-mentioned method for determining the damage assessment price of spare parts for an accident vehicle.

[0010] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: Since the damage price band information corresponding to the spare parts information is determined in advance based on several localized price characterization vectors that have similarities with each other, and a corresponding relationship between the spare parts information and the damage price band information is established, thus, based on the corresponding relationship, the target damage price band information determined by querying the corresponding relationship can provide a "price range" restriction for the damage price of the target spare parts; at the same time, since different localized feature characterization vectors corresponding to the same spare part represent the historical damage prices of the spare part in different damage assessment locations, this ensures the rationality of the "price range" given by the damage price band information determined by the localized price characterization vector. The solution provided by the embodiment of the present application is based on this reasonable price range restriction, avoiding the problem of unreasonable excessive differences in damage prices for the same spare part in different damage assessment locations. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A flowchart of a method for determining the damage assessment price of spare parts for an accident vehicle provided in an embodiment of the present application; Figure 2a A schematic diagram of the structure of the accessories price standard model in the embodiment of the present application; Figure 2b Schematic diagram of the structure and training data of the xDeepFM neural network model in the embodiment of the present application Figure 2c This is a schematic diagram of the specific structure of the “Real-time damage assessment and price quotation system for spare parts” for realizing intelligent quotation of spare parts in the embodiment of the present application; Figure 3 A schematic diagram of the specific structure of a device for determining the damage assessment price of spare parts for an accident vehicle provided in an embodiment of the present application; Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0013] Those skilled in the art may appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0014] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, which is only to describe the distinction mode adopted by the objects of the same attributes when describing in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0015] Example 1 In order to solve the problem that the damage price determination method provided by the prior art may result in unreasonable large differences in the damage price of the same spare part in different damage assessment locations, an embodiment of the present application provides a method for determining the damage price of spare parts for an accident vehicle.

[0016] The execution subject of the method may be any computing device, or may be software running on the computing device. The embodiment of the present application does not limit the computing device or software that the execution subject is.

[0017] In addition, different steps of the method may be implemented by different computing devices or software, and the embodiments of the present application do not limit which computing device or software is used to implement which step.

[0018] For ease of description, the method provided in the embodiment of the present application is described in detail below by taking the execution subject of the method as a pricing server as an example.

[0019] Please refer to the instruction manual Figure 1 A specific implementation flow chart of a method for determining the damage assessment price of accident vehicle spare parts provided in an embodiment of the present application includes the following steps: Step 11: According to the information of the target spare parts for which the damage assessment price is to be determined, by querying the pre-established correspondence between the information of different spare parts and the corresponding damage assessment price band information, the target damage assessment price band information corresponding to the information of the target spare parts is obtained; The spare parts described in the embodiments of the present application refer to products that have been completed and constitute components of the automobile, including but not limited to: wheels, air conditioners, brake discs, front fenders, headlights, taillights, engine hoods, seats, front windows, etc.

[0020] Price Zone refers to the range between the upper and lower limits of the sales price of a certain type of commodity (such as spare parts).

[0021] The damage assessment price band information in the above correspondence is determined based on a number of localized price representation vectors that are similar to each other. Different localized feature representation vectors corresponding to the same spare part represent the historical damage assessment prices of the same spare part in different damage assessment locations.

[0022] In an optional implementation, the method provided in the embodiment of the present application may include the following steps for establishing a "corresponding relationship between information of different spare parts and corresponding damage assessment price band information": First, localized feature representation vectors are obtained for representing the damage assessment prices of different spare parts at a single damage assessment location; Then, for each spare part, respectively, the following operations are performed: clustering the localized feature vectors corresponding to the spare part to obtain a cluster consisting of a number of localized feature vectors; determining the damage price band information of the spare part according to the damage price of the spare part at a single damage assessment location represented by each localized feature vector in the cluster - for example, according to the lowest damage price and the highest damage price of all prices of the spare part at a single damage assessment location represented by all localized feature vectors in the cluster, determining the damage price band of the spare part as [lowest damage price, highest damage price]; Finally, based on the information of each spare part and the determined damage assessment price band information of each spare part, a corresponding relationship between the information of different spare parts and the corresponding damage assessment price band information is established.

[0023] In an optional implementation, each localized feature characterization vector used to characterize the damage assessment price of different spare parts in a single damage assessment location may be obtained in the following manner: The Embedding vector output by the hidden layer of the fully connected neural network in the training completion state is obtained as the representation vector of each localized price.

[0024] Among them, the Embedding vector is the output of the fully connected neural network in the hidden layer based on the input sample data in the training completed state.

[0025] The input data, for example, includes: spliced ​​data consisting of spare parts description data, historical damage assessment price data of spare parts at the damage assessment location, and economic level data of the damage assessment location, and labeled.

[0026] The spare parts description data is used to describe the properties of the spare parts. The spare parts properties may include but are not limited to one or more of the following properties: spare parts name, spare parts usage, and vehicle models that the spare parts match.

[0027] The economic level data of the place where the damage is assessed may include, but is not limited to, one or more of the following data of the place where the damage is assessed: gross domestic product (GDP) data, per capita wage level data, consumer expenditure level data, urban car ownership data, commodity price index (CPI) data, etc.

[0028] The label mentioned here is the latest damage assessment price data of the spare parts at the damage assessment location.

[0029] In the embodiment of the present application, a 4-layer fully connected neural network structure (including one input layer, two hidden layers, and one output layer) can be selected for training. At the same time, the Leaky ReLU function (leaky linear rectifier function) is used as the activation function, and the RMSE (root mean square error) is used as the loss function.

[0030] In an embodiment of the present application, labeled spliced ​​data can be used as the input of the model, and the damage price data of spare parts at the damage assessment location can be predicted as the model task. The model is trained to obtain a trained deep neural network regression model.

[0031] After the model training is completed, the Embedding vector output by the second layer network of the hidden layer of the trained deep neural network regression model can be obtained as the localized price representation vector of each spare part in each damage assessment location.

[0032] The reason why the Embedding vector output by the second network layer of the hidden layer of the accessories price standard model is selected as the localized price representation vector of each spare part in each damage assessment location is because the relevant information of the label is stored in the Embedding vector output by this network layer. As can be seen from the previous introduction, the embodiment of the present application uses the latest damage assessment price data of spare parts in the damage assessment location as the label. Therefore, the Embedding vector output by this network layer stores the relevant information of the label, which is equivalent to storing the relevant information of the latest damage assessment price of spare parts in the damage assessment location. Based on this feature of the Embedding vector output by this network layer, the embodiment of the present application obtains the Embedding vector output by this network layer as the localized price representation vector of each spare part in each damage assessment location.

[0033] Step 12: based on the information affecting the damage discount rate of the target damage assessment location, predict the damage discount rate of the target spare parts at the target damage assessment location; In an optional implementation, the damage assessment discount rate of the target spare parts at the target damage assessment location may be determined in the following manner, including: Inputting information affecting the damage assessment discount rate of the target damage assessment location into the damage assessment rate prediction model; The damage rate prediction model is based on the information affecting the damage discount rate of the target damage location, and the damage discount rate of the target spare parts at the target damage location is predicted.

[0034] Among them, the damage assessment rate prediction model is pre-trained to be able to perform the task of predicting the damage assessment discount rate of the damage assessment location based on information that affects the damage assessment discount rate of the damage assessment location.

[0035] Step 13: Determine the damage price of the target spare parts at the target damage location based on the target damage price band information and the predicted damage discount rate.

[0036] By adopting the above scheme provided by the embodiment of the present application, the damage price band information corresponding to the spare parts information is determined in advance according to several localized price characterization vectors that are similar to each other, and a corresponding relationship between the spare parts information and the damage price band information is established. Therefore, based on the corresponding relationship, the target damage price band information determined by querying the corresponding relationship can provide a "price range" restriction for the damage price of the target spare parts; at the same time, since different localized feature characterization vectors corresponding to the same spare part represent the historical damage price of the spare part in different damage assessment locations, this ensures the rationality of the "price range" given by the damage price band information determined by the localized price characterization vector. The scheme provided by the embodiment of the present application is based on this reasonable price range restriction, which avoids the problem of unreasonable excessive differences in damage prices of the same spare part in different damage assessment locations.

[0037] Example 2 The following introduces a specific application example of the above method provided in an embodiment of the present application in practice.

[0038] In this application example, a Chinese insurance company wants to train an "accessory price standard model" based on the AI ​​algorithm based on the actual damage price data accumulated by the company over the years after the largest-scale negotiation with vehicle repair institutions in the insurance industry. Furthermore, in the claims process, for the damage location of the insured vehicle, the insurance company uses the "accessory price standard model" to automatically calculate the reasonable localized price standards for spare parts and to predict in real time the actual damage prices of spare parts at 4S stores and repair shops at the damage location.

[0039] Corresponding to the above goals of the insurance company, this application example is mainly divided into four parts: 1. Model training; 2. Construction of a database of localized price standards for spare parts; 3. Determination of the damage assessment discount rate; 4. Intelligent quotation for spare parts.

[0040] The following will introduce these four parts in turn.

[0041] 1. Model Training The model training process in the embodiment of the present application may include: Step 1: The model training device obtains spare parts description data respectively used to describe the properties of various spare parts that may be involved in automobile claims; The attributes of spare parts may include but are not limited to one or more of the following attributes: the name of the spare part, the purpose of the spare part, and the vehicle model that the spare part matches.

[0042] The various spare parts that may be involved in the automobile claims mentioned in the embodiments of the present application refer to products that have been completed and have become components of the automobile, including but not limited to: wheels, air conditioners, brake discs, front fenders, headlights, taillights, engine hoods, seats, front windows, etc.

[0043] Step 2: The model training device obtains the historical actual damage price data and the latest actual damage price data of various spare parts that may be involved in automobile claims in each first-level administrative region (i.e., each damage assessment location) in China, and obtains the economic level data of the first-level administrative region in the historical time period corresponding to the historical actual damage price; The historical actual damage price data and the latest actual damage price data obtained by the model training device can be collected and input into the device by a technician. For example, the technician can collect the historical actual damage price data and the latest actual damage price data of various spare parts that may be involved in automobile claims in the first-level administrative region from 4S stores and / or repair shops in each first-level administrative region.

[0044] Alternatively, the historical actual damage price data and the current latest actual damage price data may also be sent to the model training device by the staff of the 4S store and / or the repair shop using a computing device.

[0045] In the embodiment of the present application, the economic level data obtained by the model training device may include but is not limited to one or more of the data of the damage assessment location: GDP data, per capita wage level data, consumer expenditure level data, urban car ownership data, CPI data, etc.

[0046] In an embodiment of the present application, in order to enable the trained "accessory price standard model" to learn the accurate correlation between the "damage assessment price" and the "economic level" of the damage assessment location (this correlation is a complex nonlinear relationship), historical damage assessment price data and economic level data corresponding to the same historical time period range can be obtained according to pre-specified historical time period ranges.

[0047] For example, it is possible to obtain data on the historical actual damage prices of spare parts at the place of damage assessment in "2023", and on the other hand, obtain data on the economic level of the place of damage assessment in "2023".

[0048] Thus, for each combination of "a damage assessment location" and "a spare part", a data combination consisting of several "historical damage assessment price data" and "economic level data" can be obtained. Among them, the "historical damage assessment price data" and "economic level data" within the same historical time period can be regarded as one data combination.

[0049] Step 3: The model training device determines at least one of the following data: price range, price point and price change trend of the spare parts at the damage assessment location based on the historical damage assessment price data of the spare parts at the damage assessment location, as the processed historical damage assessment price data; Price Zone refers to the range between the upper and lower limits of the sales price of a certain type of commodity (such as spare parts).

[0050] Price point refers to the standard retail price used by marketers for several products that have slightly different wholesale costs but share the same level of consumer perceived value (such as the windshield glass component type, where there are actually multiple windshield glass products with slightly different qualities).

[0051] Price trend refers to the basic direction of price change over a period of time. For example, price change trends can be divided into three basic types: upward trend, downward trend and sideways trend (also called horizontal trend or oscillating trend).

[0052] Step 4: The model training device executes for each combination of "a damage assessment location" and "a spare part": first, the spare part description data of the spare part, the historical damage assessment price data of the spare part after processing at the damage assessment location, and the economic level data of the damage assessment location are spliced ​​to obtain spliced ​​data; then, the latest damage assessment price data of the spare part at the damage assessment location is used as a label, and the obtained spliced ​​data is labeled; each labeled spliced ​​data is used as a training sample to train the deep neural network regression model to be trained, and a trained deep neural network regression model is obtained as a "parts price standard model".

[0053] The specific implementation process of step 4 may include the following sub-steps: Sub-step 41: constructing a deep neural network regression model to be trained; For example, you can choose a 4-layer fully connected neural network structure (including one input layer, two hidden layers, and one output layer). At the same time, use the Leaky ReLU function (leaky linear rectifier function) as the activation function and RMSE (root mean square error) as the loss function.

[0054] Sub-step 42: Use each training sample to train the deep neural network regression model to be trained until the trained converged model is obtained as the “accessory price standard model”.

[0055] In the process of training the deep neural network regression model to be trained in sub-step 42, the model parameters can be optimized through the back propagation algorithm with the goal of minimizing the function value of the loss function.

[0056] At this point, the first part "model training" is completed.

[0057] 2. Construction of a database of localized price standards for spare parts The second part mainly includes the following steps: Step a: After the model training is completed, the model training device obtains the Embedding vector output by the second layer network of the hidden layer of the spare parts price standard model as the localized price representation vector of each spare part in each damage assessment location; The reason why the Embedding vector output by the second network layer of the hidden layer of the accessories price standard model is selected as the localized price representation vector of each spare part in each damage assessment location is because the relevant information of the label is stored in the Embedding vector output by this network layer. As can be seen from the previous introduction, the embodiment of the present application uses the latest damage assessment price data of spare parts in the damage assessment location as the label. Therefore, the Embedding vector output by this network layer stores the relevant information of the label, which is equivalent to storing the relevant information of the latest damage assessment price of spare parts in the damage assessment location. Based on this feature of the Embedding vector output by this network layer, the embodiment of the present application obtains the Embedding vector output by this network layer as the localized price representation vector of each spare part in each damage assessment location.

[0058] Step b: The model training device stores the localized price representation vector in Faiss; Faiss is Facebook's open source AI similarity search tool, which has a vector database that can store vectors. Localized price representation vectors can be stored in this vector database.

[0059] Step c: The model training device uses a localized price representation vector of each spare part as a search vector, and retrieves localized price representation vectors (called vector sets) corresponding to the spare part and similar to the search vector from the Faiss vector database; By executing step c, a vector set corresponding to each component can be obtained.

[0060] The retrieval principle of Faiss for vector databases is: based on efficient vector indexing technology, find the vector most similar to the query vector in the vector database. Faiss achieves fast retrieval by building index structures, such as quantitative indexes or tree structure indexes. During the index building process, it maps data points to a compact representation, which helps reduce storage space and increase retrieval speed. When searching, Faiss uses these index structures to quickly screen out the closest candidate vectors, and then determines the final retrieval results through precise similarity calculations. This retrieval process not only improves efficiency, but also significantly reduces the consumption of computing resources while maintaining high accuracy.

[0061] Based on this retrieval principle, taking the spare part "Passat car headlight" as an example, in step c, the model training device can use a localized price representation vector of the spare part as a retrieval vector to retrieve several candidate vectors that are relatively similar to the retrieval vector.

[0062] Among them, a localized price representation vector of the spare part can be directly input by a technician to the model training device; or, the technician can input the name of the spare part, such as "Passat car headlight", and the name of the damage assessment place, such as "Hebei" into the model training device, and then the model training device will convert the acquired spare part name and the name of the damage assessment place into a corresponding vector as a localized price representation vector of the spare part.

[0063] As mentioned above, a single localized price representation vector (i.e., Embedding vector) stores the latest damage assessment price of a single spare part in a single damage assessment location. Therefore, the candidate vectors retrieved generally meet the following conditions: 1. They correspond to the spare part "Passat car headlight"; 2. The latest damage assessment price of the corresponding spare part in a single damage assessment location is similar to the latest damage assessment price of the spare part in a single damage assessment location corresponding to the search vector.

[0064] Step d: for each vector set corresponding to each spare part retrieved, respectively perform: using the DBSCAN clustering algorithm, perform clustering operations on the localized price representation vectors in the vector set, so as to obtain each cluster corresponding to each spare part; The DBSCAN clustering algorithm is a density-based spatial clustering algorithm that divides areas with sufficiently high density into clusters and can find clusters of any shape in a noisy spatial database. The core idea of ​​the algorithm is: for each data point, if there are a sufficient number of other points within a certain range around it, the point is considered a core point; otherwise, it is a boundary point or a noise point. The neighborhood of the core point contains at least the minimum number of points (MinPts), while the boundary point is located in the neighborhood of the core point but does not meet the core point conditions. In this way, the DBSCAN algorithm can identify high-density areas and cluster the points in these areas, while treating low-density areas as noise points. In this method, the DBSCAN clustering algorithm is used to cluster the Embedding vectors to identify clusters of spare parts with similar damage assessment prices, thereby providing a basis for determining the damage assessment prices of accident vehicle spare parts.

[0065] In an embodiment of the present application, the parameter values ​​of the two parameters "neighborhood radius (eps)" and "minimum number of points (MinPts)" can be set according to actual needs. In the DBSCAN algorithm, the data point density is defined by the number of data points in the neighborhood within a specified radius of a given data point. Under the definition of density, the DBSCAN algorithm divides data points into three categories: core points, boundary points, and noise points. If a target data point, that is, a localized price representation vector, contains at least the number of MinPts data points in its eps-neighborhood, then the target data point is regarded as a core point. The DBSCAN algorithm selects core points, and for each core point, connects all points in the eps-neighborhood of the core point (including other core points) to form a cluster, and points belonging to the same category are aggregated together.

[0066] As mentioned above, the localized price representation vector stores the relevant information of the latest damage appraisal price of spare parts at the damage appraisal location. Therefore, the cluster corresponding to the same spare part obtained by clustering the localized price representation vectors in the vector set satisfies: "The difference between the latest damage appraisal prices at the damage appraisal location represented by each localized price representation vector in the same cluster is small, and the corresponding difference between "the latest damage appraisal prices at the damage appraisal location represented by each localized price representation vector in different clusters" is large.

[0067] Based on the above characteristics of the cluster, the localized price representation vectors in the same cluster can refer to each other. Therefore, further, the price band corresponding to the cluster can be determined according to the localized price representation vectors in the same cluster, as the price band (price range) of the latest damage assessment price of the corresponding spare parts in each damage assessment location. The specific implementation process of determining the price band corresponding to the cluster according to the localized price representation vectors in the same cluster includes: 1. First, sort all the localized price representation vectors in the selected cluster to ensure the continuity of the price. 2. Then, calculate the average value of the damage assessment price represented by each vector in the cluster as the central price point of the cluster. 3. Then, according to the distribution of prices in the cluster, determine the upper and lower limits of the price band, that is, the width of the price band; 4. Finally, the calculated central price point and price band width can be used as the price band information of the cluster for subsequent intelligent quotation of spare parts.

[0068] It should be noted that, to facilitate subsequent inquiries, in the embodiment of the present application, after determining the price band information of the latest damage assessment price of spare parts in several damage assessment locations, a correspondence between the identification of the spare parts and the corresponding price band information can be established - for example, a correspondence between the "Passat car headlights" and the price band information of the latest damage assessment price of the spare parts in the damage assessment location can be established.

[0069] The corresponding relationship may be stored in a database, which may be referred to as a spare parts localization price standard database.

[0070] III. Determination of the Discount Rate for Damage Assessment Since the actual damage assessment price of spare parts will be affected by multiple factors such as time, region, vehicle brand, model and year, repair shop type, price plan, and damage assessment premium ratio, the existing method of setting a fixed damage assessment discount rate between insurance companies and 4S stores (or repair shops) has major flaws.

[0071] In an embodiment of the present application, a deep neural network is used to construct a regression model (called a damage discount rate prediction model) to fit the relationship between the actual damage discount rate of spare parts and multiple factors that affect the actual damage discount rate, replacing the current method of setting a fixed damage discount rate in the industry.

[0072] The damage discount rate prediction model can utilize the long-term historical characteristics related to each 4S store (or repair shop) and spare parts, as well as the short-term real-time characteristics that may affect the actual damage price of spare parts, to predict in real time the reasonable damage discount rate of any spare part of a 4S store (or repair shop) at a certain damage assessment point in the damage assessment quotation business process.

[0073] In a specific example, the embodiment of the present application uses a regression model built based on the xDeepFM neural network model as the damage assessment discount rate prediction model to estimate the actual damage assessment price of a certain spare part in a 4S store (or repair shop) at a certain point in time. Among them, the xDeepFM neural network model is a deep learning model for recommendation systems, which aims to improve the accuracy of predictions by combining explicit and implicit feature interactions.

[0074] Please refer to the instruction manual Figure 2b The xDeepFM neural network model includes an embedding layer and three modules: a linear layer, a fully connected layer (Plain DNN), and a compressed sensing layer (Compressed Interaction Network, CIN). The linear layer processes sparse 01 vectors, such as Figure 2b The fully connected layer (Plain DNN) processes the dense vector after embedding, and the compressed sensing layer (CIN) is the core of xDeepFM, which is used to explicitly construct finite-order feature combinations.

[0075] The workflow of xDeepFM includes: after the original data is processed by the embedding layer, it is divided into three parts and input into Linear, Plain DNN and CIN. Linear directly uses the original sparse vector, Plain DNN processes dense vector, and CIN performs explicit high-order feature interaction. Finally, the results of these three parts are fused and the final prediction result is output through the sigmoid function. Among them, the CIN network is the most core part. The CIN network has a vector wise form and uses a convolutional neural network to mine potential nonlinear relationships based on high-order feature interactions, which can more accurately fit the relationship between features and spare parts damage discount rates.

[0076] In an embodiment of the present application, the original data includes: original data having long-term historical characteristics related to each 4S store (or repair shop) and spare parts (first original data), and original data having short-term real-time characteristics related to each 4S store (or repair shop) and spare parts (second original data).

[0077] Among them, the long-term historical characteristics refer to the characteristics that are relatively less affected by the specific claims case situation and time changes, and are applicable as the basis for determining the discount rate of spare parts damage assessment in different claims cases and time points. For example, it includes but is not limited to one or more of these characteristics: the price type of spare parts (market adjustment price, government guidance price, government pricing) characteristics, the local (target damage assessment location) reference price characteristics of spare parts, the damage assessment price characteristics of spare parts by different repair shops of the same type, the third-party reference price characteristics, etc. The corresponding data including these characteristics include one or more of these data: the price type of spare parts (market adjustment price, government guidance price, government pricing), the local (target damage assessment location) reference price of spare parts, the damage assessment price of spare parts by different repair shops of the same type, the third-party reference price, etc.

[0078] The short-term real-time features are those that are relatively greatly affected by the specific claims cases and time changes and are not suitable as the basis for determining the discount rate of spare parts damage assessment in different claims cases or time points. For example, they include but are not limited to one or more of these features: damage assessment method features, damage assessment point features, damage assessor features, etc. The corresponding data including these features include one or more of these data: damage assessment method, damage assessment point, damage assessor, etc.

[0079] In the embodiment of the present application, during the training phase of the damage assessment discount rate prediction model, pytorch, a deep learning framework that is a collection of tools for building, training, and deploying deep neural network models, can be used to complete the training of the damage assessment discount rate prediction model. When training the model, the sample data can be labeled sample data: the sample data can be, for example, multiple factors that affect the actual damage assessment discount rate; the sample label can specifically be the damage assessment discount rate for spare parts.

[0080] After the model training is completed, the trained damage discount rate prediction model can be converted into an Open Neural Network Exchange (ONNX) format model for subsequent prediction of real-time damage discount rates. ONNX is a standard format for representing deep learning models, allowing models to be exchanged between different machine learning frameworks and tools.

[0081] 4. Intelligent quotation of spare parts In the present application embodiment, Figure 2c As shown, it is a schematic diagram of the specific structure of the "Real-time Damage Assessment Price Quotation System for Spare Parts" for realizing intelligent quotation of spare parts. The system includes a big data application platform and a damage assessment platform.

[0082] Multiple damage assessment platforms can refer to the general term for multiple damage assessment platforms in different damage assessment locations, all of which can be software systems that at least have the function of "vehicle spare parts damage assessment price query and display".

[0083] A big data application platform, in which a "parts damage price prediction model" is run, which may include a trained "damage discount rate prediction model" and a "spare parts localized price standard database". The big data application platform is specifically used to perform the following operations: 1. According to the identification of the spare parts (such as the headlight of a Passat car) contained in the model calling parameters sent by the damage assessment platform, call the "parts damage assessment price prediction model" to execute: query the price band information corresponding to the identification of the spare parts from the spare parts local price standard database constructed and deployed in the big data application platform in advance using the method described in the second part "Construction of the spare parts local price standard database" as described above; Among them, the model calling parameters sent by the damage assessment platform can be input into the damage assessment platform by personnel who use the damage assessment platform to query the real-time damage assessment prices of spare parts.

[0084] Model calling parameters can generally be divided into two categories. One category is the parameters used as query conditions for localized price standard data of spare parts, such as the identification of spare parts; the other category is the parameters used as query conditions for damage assessment discount rate. These parameters can specifically be multiple factors that affect the actual damage assessment discount rate of the damage assessment location (such as Hebei), such as but not limited to the identification of spare parts, the time point of damage assessment, the identification of the damage assessment location, the brand, model, year and age of the damaged vehicle, the type of repair shop, etc.

[0085] 2. According to the parameters used as the damage assessment discount rate query conditions contained in the model call parameters, call the "Accessory Damage Assessment Price Prediction Model" to execute: from the "Damage Assessment Discount Rate Prediction Model" trained and deployed in the big data application platform in advance using the method described in the third part "Determination of Damage Assessment Discount Rate" as described above, so as to obtain the "damage assessment discount rate prediction model" using these parameters as input to predict the "real-time damage assessment discount rate of the spare part at the damage assessment location" - that is, Figure 2c "Accessory price discount rate Y" in; 3. The "parts damage assessment price prediction model" determines the real-time damage assessment price information of the spare part at the damage assessment location based on the queried price band information corresponding to the identification of the spare part and the obtained "real-time damage assessment discount rate of the spare part at the damage assessment location"; For example, the real-time damage assessment center price p of the damage assessment location can be calculated according to the central price point in the price band information and the calculation method of "central price point × real-time damage assessment discount rate of the spare part at the damage assessment location". Furthermore, according to the price band width m in the price band information, [p-(2 / m), p+(2 / m)] can be calculated as the real-time damage assessment price range of the damage assessment location. The real-time damage assessment center price p and the real-time damage assessment price range of the damage assessment location can be used together as the real-time damage assessment price information of the spare part at the damage assessment location.

[0086] 4. The big data application platform obtains the real-time damage price information of the spare part at the damage location determined by the "parts damage price prediction model", and sends the real-time damage price information to the damage platform for display.

[0087] For example, the damage assessment platform can adopt Figure 2c The “intelligent quotation display” method shown displays the real-time damage assessment price.

[0088] By adopting the solution provided in the embodiment of the present application, by deploying the above-mentioned "parts damage assessment price prediction model" on the big data application platform, and supporting different damage assessment platforms to call the model by sending model call parameters to the big data application platform, it can support various 4S stores (repair shops) to provide smart quotes in seconds for the damage assessment prices of tens of millions of parts. Among them, relative to the physical location of the damage assessment platform, the big data application platform includes but is not limited to a server in the cloud, or it can be a graphics processing unit (GPU) of the local device where the damage assessment platform runs, etc.

[0089] Example 3 Based on the same inventive concept as the above embodiments, in order to solve the problem that the damage price determination method provided by the prior art may result in unreasonable differences in the damage prices of the same spare parts in different damage assessment locations, the present embodiment provides a method such as Figure 3 The device for determining the damage price of spare parts of an accident vehicle shown in the figure includes the following functional units: The price band determination unit 31 is used to obtain the target damage assessment price band information corresponding to the information of the target spare part by querying the pre-established correspondence relationship between the information of different spare parts and the corresponding damage assessment price band information according to the information of the target spare part for which the damage assessment price is to be determined; the damage assessment price band information in the correspondence relationship is determined according to a number of localized price representation vectors having similarities with each other; different localized feature representation vectors corresponding to the same spare part represent the historical damage assessment prices of the same spare part in different damage assessment locations; A discount rate prediction unit 32, configured to predict the damage discount rate of the target spare part at the target damage assessment location based on information affecting the damage discount rate at the target damage assessment location; The damage assessment price determination unit 33 is used to determine the damage assessment price of the target spare part at the target damage assessment location according to the target damage assessment price band information and the predicted damage assessment discount rate.

[0090] In an optional implementation, the device provided in the embodiment of the present application may further include the following functional units: A vector acquisition unit, used to acquire localized feature characterization vectors respectively used to characterize the damage assessment prices of different spare parts in a single damage assessment location; The clustering unit is used to respectively perform, for each spare part, the following steps: clustering the localized feature vectors corresponding to the spare part to obtain a cluster consisting of a plurality of the localized feature vectors; determining the damage price band information of the spare part according to the damage price of the spare part in a single damage assessment location represented by each localized feature vector in the cluster; The corresponding relationship establishing unit is used to establish the corresponding relationship according to the information of each spare part and the determined damage assessment price band information of each spare part.

[0091] In an optional implementation, the vector acquisition unit may be specifically used for: The Embedding vector output by the hidden layer of the fully connected neural network in the training completion state is obtained as the localized price representation vector.

[0092] The embedding vector is the output of the fully connected neural network in the training completed state, based on the input sample data, in the hidden layer; The input data includes: spliced ​​data consisting of spare parts description data, historical damage assessment price data of spare parts at the damage assessment location and economic level data of the damage assessment location, and marked with a label; the label is the latest damage assessment price data of spare parts at the damage assessment location.

[0093] In an optional implementation, the discount rate prediction unit 32 may be specifically used for: Inputting the information affecting the damage assessment discount rate of the target damage assessment location into a damage assessment rate prediction model; the damage assessment rate prediction model is pre-trained to be able to perform a task of predicting the damage assessment discount rate of the damage assessment location based on the information affecting the damage assessment discount rate of the damage assessment location; The damage assessment rate prediction model is used to obtain the damage assessment discount rate of the target spare parts at the target damage assessment location based on the information affecting the damage assessment discount rate at the target damage assessment location.

[0094] In an optional implementation, the information that affects the damage assessment discount rate at the target damage assessment location includes: long-term historical characteristics and short-term real-time characteristics related to the repairer and spare parts at the target damage assessment location.

[0095] By using the above-mentioned device provided by the embodiment of the present application, the damage price band information corresponding to the spare parts information is determined in advance according to several localized price characterization vectors that are similar to each other, and a corresponding relationship between the spare parts information and the damage price band information is established. Therefore, based on the corresponding relationship, the target damage price band information determined by querying the corresponding relationship can provide a "price range" restriction for the damage price of the target spare parts; at the same time, since different localized feature characterization vectors corresponding to the same spare part represent the historical damage price of the spare part in different damage assessment locations, this ensures the rationality of the "price range" given by the damage price band information determined by the localized price characterization vector. The solution provided by the embodiment of the present application is based on this reasonable price range restriction, which avoids the problem of unreasonable excessive differences in damage prices of the same spare part in different damage assessment locations.

[0096] Example 4 Based on the same inventive concept as the previous embodiments, the embodiments of the present application also provide an electronic device to solve the problem that the damage price determination method provided by the prior art may result in unreasonable large differences in damage prices for the same spare part in different damage assessment locations.

[0097] like Figure 4 As shown, the electronic device includes: a memory 41 and a processor 42. The memory 41 can be configured to store various other data to support operations on the electronic device. Examples of these data include instructions for any application or method for operating on the electronic device. The memory 41 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0098] The processor 42 is coupled to the memory 41 and is used to execute the program stored in the memory 41 to execute the method for determining the damage assessment price of accident vehicle spare parts described in the embodiment of the present application.

[0099] When the processor 42 executes the program in the memory 41, in addition to the above functions, it can also realize other functions, and the details can be referred to the description of the previous embodiments.

[0100] Further, such as Figure 4As shown, the electronic device also includes: a display 44, a communication component 43, a power component 45, an audio component 46 and other components. Figure 4 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 4 Components shown.

[0101] Accordingly, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a computer, the steps or functions of the methods provided in the above embodiments can be implemented.

[0102] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0103] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for determining the damage price of accident vehicle spare parts, characterized in that: The method comprises: According to the information of the target spare parts for which the damage assessment price is to be determined, the target damage assessment price band information corresponding to the information of the target spare parts is obtained by querying the correspondence relationship between the information of different spare parts and the corresponding damage assessment price band information established in advance; the damage assessment price band information in the correspondence relationship is determined based on a number of localized price representation vectors that are similar to each other; different localized feature representation vectors corresponding to the same spare part represent the historical damage assessment prices of the same spare part in different damage assessment locations; Based on information affecting the damage assessment discount rate of the target damage assessment location, predicting the damage assessment discount rate of the target spare part at the target damage assessment location; The damage assessment price of the target spare parts at the target damage assessment location is determined based on the target damage assessment price band information and the predicted damage assessment discount rate.

2. The method according to claim 1, characterized in that The method further comprises: Obtaining localized feature representation vectors respectively used to represent the damage assessment prices of different spare parts in a single damage assessment location; For each spare part, respectively perform: clustering the localized feature vector corresponding to the spare part to obtain a cluster composed of a plurality of the localized feature vectors; determining the damage price band information of the spare part according to the damage price of the spare part in a single damage assessment location represented by each localized feature vector in the cluster; The corresponding relationship is established based on the information of each spare part and the determined damage price band information of each spare part.

3. The method according to claim 2, characterized in that Obtain each localized feature representation vector used to represent the damage assessment price of different spare parts in a single damage assessment location, including: Obtaining the Embedding vector output by the hidden layer of the fully connected neural network in the training completed state as the localized price representation vector; The embedding vector is the output of the fully connected neural network in the training completed state, based on the input sample data, in the hidden layer; The input data includes: spliced ​​data consisting of spare parts description data, historical damage assessment price data of spare parts at the damage assessment location and economic level data of the damage assessment location, and marked with a label; the label is the latest damage assessment price data of spare parts at the damage assessment location.

4. The method according to any one of claims 1 to 3, characterized in that: Based on information affecting the damage assessment discount rate of the target damage assessment location, predicting the damage assessment discount rate of the target spare part at the target damage assessment location includes: Inputting the information affecting the damage assessment discount rate of the target damage assessment location into a damage assessment rate prediction model; the damage assessment rate prediction model is pre-trained to be able to perform a task of predicting the damage assessment discount rate of the damage assessment location based on the information affecting the damage assessment discount rate of the damage assessment location; The damage assessment rate prediction model is used to obtain the damage assessment discount rate of the target spare parts at the target damage assessment location based on the information affecting the damage assessment discount rate at the target damage assessment location.

5. The method according to claim 4, characterized in that The information affecting the damage assessment discount rate of the target damage assessment location includes: long-term historical characteristics and short-term real-time characteristics related to the repair party and spare parts at the target damage assessment location.

6. A device for determining the damage price of accident vehicle spare parts, characterized in that: The device comprises: The price band determination unit is used to obtain the target damage assessment price band information corresponding to the information of the target spare part by querying the pre-established correspondence relationship between the information of different spare parts and the corresponding damage assessment price band information according to the information of the target spare part for which the damage assessment price is to be determined; the damage assessment price band information in the correspondence relationship is determined according to a number of localized price characterization vectors having similarities with each other; different localized feature characterization vectors corresponding to the same spare part characterize the historical damage assessment prices of the same spare part in different damage assessment locations; A discount rate prediction unit, configured to predict the damage discount rate of the target spare part at the target damage assessment location based on information affecting the damage discount rate at the target damage assessment location; The damage assessment price determination unit is used to determine the damage assessment price of the target spare part at the target damage assessment location based on the target damage assessment price band information and the predicted damage assessment discount rate.

7. The device according to claim 6, characterized in that The device also includes: A vector acquisition unit, used to acquire localized feature characterization vectors respectively used to characterize the damage assessment prices of different spare parts in a single damage assessment location; The clustering unit is used to respectively perform, for each spare part, the following steps: clustering the localized feature vectors corresponding to the spare part to obtain a cluster consisting of a plurality of the localized feature vectors; determining the damage price band information of the spare part according to the damage price of the spare part in a single damage assessment location represented by each localized feature vector in the cluster; The corresponding relationship establishing unit is used to establish the corresponding relationship according to the information of each spare part and the determined damage assessment price band information of each spare part.

8. An electronic device, characterized in that: include: Memory and processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to execute the method for determining the damage assessment price of accident vehicle spare parts as described in any one of claims 1 to 5.

9. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer, can implement the method for determining the damage assessment price of spare parts for an accident vehicle as described in any one of claims 1 to 5.

10. A computer program product, characterized in that The computer program product stores instructions, which, when executed by a computer, enable the computer to implement the method according to any one of claims 1 to 10.

Images