Training method and device for repair shop recommendation model of insurance accident vehicle

Abstract

The embodiment of the invention provides a training method and device for a repair shop recommendation model of an insurance accident vehicle, and the method comprises the steps: creating the repair shop recommendation model, setting two sub-models in the repair shop recommendation model, carrying out the feature extraction of historical repair pushing data, obtaining first training sample data and second training sample data, and carrying out the feature extraction of the first training sample data and the second training sample data; inputting training sample data into the two sub-models, predicting the success rate of recommending a comprehensive repair factory through the first sub-model, predicting the success rate of recommending all types of repair factories through the second sub-model, and adjusting the parameters of the first sub-model and the second sub-model to obtain a trained repair factory recommendation model. The repair factory is pushed to the insurance customer more accurately, the satisfaction degree of the customer is improved, and the claim settlement cost is reduced.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more particularly to a training method and device for a repair shop recommendation model for insurance accident vehicles. Background Art

[0002] When a user files a claim with an insurance company after a vehicle accident, the insurance company will send the accident vehicle to a repair shop for damage assessment and repairs, and will also send the user information about the repair shop recommended by the insurance company. The user will then decide whether to choose the repair shop recommended by the insurance company based on the push results.

[0003] In the prior art, the repair shops pushed by insurance companies are divided into comprehensive repair shops and 4S shop repair shops. Insurance companies build a repair management system, configure repair rules, write the repair rules into the system, and realize the push of repair shops.

[0004] However, in the prior art, the repair shop actually selected by the user is often not the one recommended by the repair management system. The low success rate of repair recommendation will reduce customer satisfaction. When the user is not satisfied with the repair shop recommended by the insurance company, he or she will usually give priority to sending the accident vehicle to a 4S shop for repair. The repair fee of a 4S shop is usually higher than that of a comprehensive repair shop. If all accident vehicles are sent to a 4S shop for repair, the insurance company's claim settlement cost will be greatly increased. Summary of the invention

[0005] The embodiment of the present application provides a training method and device for a repair shop recommendation model for insurance accident vehicles, so as to solve the problems of high claims settlement costs and reduced customer satisfaction of insurance companies existing in the prior art.

[0006] In a first aspect, an embodiment of the present application provides a training method for a repair shop recommendation model for an insurance accident vehicle, wherein the architecture of the repair shop recommendation model includes a first sub-model and a second sub-model; wherein the first sub-model is used to predict the success rate of recommending a comprehensive repair shop, and the second sub-model is used to predict the success rate of recommending all types of repair shops;

[0007] The first sub-model includes: a first embedding layer, a first feature concatenation layer and a first fully connected layer; the second sub-model includes: a second embedding layer, a second feature concatenation layer and a second fully connected layer;

[0008] The method comprises:

[0009] Extract a preset number of historical recommended repair data from the repair system;

[0010] Perform feature extraction on each piece of historical recommended repair data to obtain first training sample data and second training sample data;

[0011] The first training sample data includes the accident vehicle feature domain, the comprehensive repair shop feature domain, the vehicle-shop cross feature domain and the labeling success rate of the recommended comprehensive repair shop;

[0012] The second training sample data includes the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains and the labeling success rate of all types of repair shops;

[0013] Input the accident vehicle feature domain, the comprehensive repair shop feature domain, and the vehicle-shop cross feature domain in each first training sample data into the first sub-model, and output the predicted success rate of the comprehensive repair shop after being processed by the first embedding layer, the first feature concatenation layer, and the first fully connected layer in sequence;

[0014] Input the accident vehicle feature domain, various types of repair shop feature domains, and vehicle-shop cross feature domains in each second training sample data into the second sub-model, and output the predicted success rate of all types of repair shops after being processed by the second embedding layer, the second feature concatenation layer, and the second fully connected layer in sequence;

[0015] Adjusting the model parameters of the first sub-model and the second sub-model according to the prediction success rate of the comprehensive repair shop and the marking success rate of the comprehensive repair shop, as well as the prediction success rate of all types of repair shops and the marking success rate of all types of repair shops;

[0016] When the prediction accuracy of the first sub-model and the second sub-model meets the preset standard, the training is stopped to obtain a trained repair shop recommendation model.

[0017] In one possible implementation, the first embedding layer of the first sub-model and the second embedding layer of the second sub-model are shared embedding layers; the first embedding layer of the first sub-model is used to input the first training sample data and the second training sample data; the second embedding layer of the second sub-model is used to input the first training sample data and the second training sample data.

[0018] In a possible implementation, before adjusting the model parameters of the first sub-model and the second sub-model according to the prediction success rate of the comprehensive repair shop and the marking success rate of the comprehensive repair shop, as well as the prediction success rate of all types of repair shops and the marking success rate of all types of repair shops, it also includes: creating an auxiliary target success rate according to the prediction success rate of the comprehensive repair shop and the prediction success rate of all types of repair shops; training the second sub-model according to the auxiliary target success rate, and calculating and generating a loss function value according to the prediction success rate of the comprehensive repair shop and the prediction success rate of all types of repair shops; adjusting the second sub-model according to the loss function value to generate a trained second sub-model.

[0019] In a possible implementation, the formula for calculating and generating the loss function is:

[0020]

[0021] In the formula, x i Represents the input features; y i Represents the output prediction result; represents a parameter set of the second sub-model; represents the parameter set of the first sub-model; p3 represents the auxiliary target success rate.

[0022] In a possible implementation manner, the repair shop recommendation model is:

[0023] argmax(p1,p2)=f(X1,X2,…,X n )

[0024] Where p1 represents the prediction success rate of all types of repair shops; p2 represents the prediction success rate of comprehensive repair shops; f() represents the linear relationship between p1 and p2; X n Indicates the factors that affect the success rate of the repair.

[0025] In a second aspect, an embodiment of the present application provides a training device for a repair shop recommendation model for insurance accident vehicles, wherein the architecture of the repair shop recommendation model includes a first sub-model and a second sub-model; wherein the first sub-model is used to predict the success rate of recommending a comprehensive repair shop, and the second sub-model is used to predict the success rate of recommending all types of repair shops;

[0026] The first sub-model includes: a first embedding layer, a first feature concatenation layer and a first fully connected layer; the second sub-model includes: a second embedding layer, a second feature concatenation layer and a second fully connected layer;

[0027] The device comprises:

[0028] The first extraction module is used to extract a preset number of historical recommended repair data from the repair system;

[0029] A second extraction module is used to extract features from each piece of historical recommended repair data to obtain first training sample data and second training sample data;

[0030] The first training sample data includes the accident vehicle feature domain, the comprehensive repair shop feature domain, the vehicle-shop cross feature domain and the labeling success rate of the recommended comprehensive repair shop;

[0031] The second training sample data includes the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains and the labeling success rate of all types of repair shops;

[0032] A first input module is used to input the accident vehicle feature domain, the comprehensive repair shop feature domain and the vehicle-shop cross feature domain in each first training sample data into the first sub-model, and output the predicted success rate of the comprehensive repair shop after being processed by the first embedding layer, the first feature concatenation layer and the first fully connected layer in sequence;

[0033] A second input module is used to input the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains in each second training sample data into the second sub-model, and output the predicted success rate of all types of repair shops after being processed by the second embedding layer, the second feature concatenation layer and the second fully connected layer in sequence;

[0034] an adjustment module, configured to adjust the model parameters of the first sub-model and the second sub-model according to the prediction success rate of the comprehensive repair shop and the marking success rate of the comprehensive repair shop, and the prediction success rate of all types of repair shops and the marking success rate of all types of repair shops;

[0035] The stopping module is used to stop training to obtain a trained repair shop recommendation model when the prediction accuracy of the first sub-model and the second sub-model meets a preset standard.

[0036] In one possible implementation, the first embedding layer of the first sub-model and the second embedding layer of the second sub-model are shared embedding layers; the first embedding layer of the first sub-model is used to input the first training sample data and the second training sample data; the second embedding layer of the second sub-model is used to input the first training sample data and the second training sample data.

[0037] In a third aspect, an embodiment of the present application provides a training device for a repair shop recommendation model for an insurance accident vehicle, including:

[0038] at least one processor and memory;

[0039] The memory stores computer-executable instructions;

[0040] The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the above first aspect and / or various possible implementations of the first aspect.

[0041] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementations of the first aspect.

[0042] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.

[0043] The training method and device for the repair shop recommendation model for insurance accident vehicles provided in the embodiment of the present application, by creating a repair shop recommendation model, setting two sub-models in the repair shop recommendation model, performing feature extraction on historical recommended repair data to obtain first training sample data and second training sample data, inputting the training sample data into the two sub-models, predicting the success rate of recommending comprehensive repair shops through the first sub-model, predicting the success rate of recommending all types of repair shops through the second sub-model, and adjusting the parameters of the first sub-model and the second sub-model to obtain a trained repair shop recommendation model. Compared with the prior art, the repair shops can be pushed to insurance customers more accurately, thereby improving customer satisfaction and reducing claims costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0045] Figure 1 A schematic diagram of the system structure of a computer device provided in an embodiment of the present application;

[0046] Figure 2 A flowchart of a training method for a repair shop recommendation model for insurance accident vehicles provided in this application;

[0047] Figure 3 A schematic diagram of the recommended model of the repair shop provided for this application;

[0048] Figure 4 A schematic diagram of the structure of a training device for a repair shop recommendation model for insurance accident vehicles provided in this application;

[0049] Figure 5 A schematic diagram of the structure of the training device for the repair shop recommendation model for insurance accident vehicles provided in this application.

[0050] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0051] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0052] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0053] When a user files a claim with an insurance company after an accident, the insurance company will push the accident vehicle to a repair shop for vehicle damage assessment and repair, and push the information of the repair shop recommended by the insurance company to the user. The user decides whether to choose the repair shop recommended by the insurance company based on the push results. In the prior art, the repair shops pushed by insurance companies are divided into comprehensive repair shops and 4S shop repair shops. Insurance companies build a repair management system, configure the repair rules, write the repair rules into the system, and push the repair shop. However, in the prior art, the repair shop actually selected by the user is often not the repair shop recommended by the repair management system, and the low success rate of repair recommendation will reduce customer satisfaction. When the user is not satisfied with the repair shop recommended by the insurance company, it will usually give priority to sending the accident vehicle to the 4S shop for repair. The repair cost of the 4S shop repair shop is usually higher than that of the comprehensive repair shop. If all accident vehicles are pushed to the 4S shop repair shop for repair, the insurance company's claim cost will be greatly increased.

[0054] In order to solve the above technical problems, the embodiment of the present application proposes the following technical concept: the inventor considers creating a repair shop recommendation model for insurance accident vehicles, sets two sub-models in the repair shop recommendation model, uses the first sub-model to predict the success rate of recommending comprehensive repair shops, and uses the second sub-model to predict the success rate of recommending all types of repair shops. Considering the feature extraction of historical recommended repair data, the first training sample and the second training sample are obtained, and the training samples are respectively input into the two sub-models, and the predicted success rate of the comprehensive repair shop and the predicted success rate of all types of repair shops are obtained through the embedding layer, the feature concatenation layer and the fully connected layer in sequence. The repair shop recommendation model is adjusted by the predicted success rate and the marked success rate in the training sample data to obtain a trained repair shop recommendation model. The repair shop is predicted by the trained repair shop recommendation model, which solves the problems of increased claims settlement costs and reduced customer satisfaction of insurance companies in the prior art. The following is a detailed description using a detailed embodiment.

[0055] Figure 1 A schematic diagram of the system structure of a computer device provided in an embodiment of the present application. Figure 1 As shown, the computer device includes: a receiving device 101, a processor 102 and a display device 103.

[0056] It is understandable that the structure illustrated in the embodiment of the present application does not constitute a specific limitation on the object identification method. In other feasible implementations of the present application, the above architecture may include more or fewer components than shown in the figure, or combine certain components, or split certain components, or arrange the components differently, which can be determined according to the actual application scenario and is not limited here. Figure 1 The components shown may be implemented in hardware, software, or a combination of software and hardware.

[0057] In a specific implementation process, the receiving device 101 may be an input / output interface or a communication interface, and may extract a preset number of historical recommended repair data.

[0058] The processor 102 may output the predicted success rate of the comprehensive repair shop and the predicted success rate of all types of repair shops.

[0059] The display device 103 can be used to display the predicted success rate of the above-mentioned comprehensive repair shop and the predicted success rate of all types of repair shops.

[0060] The display device may also be a touch display screen, which is used to receive user instructions while displaying the above-mentioned content, so as to realize operational interaction with the user.

[0061] It should be understood that the above-mentioned processor can be implemented by the processor reading instructions in the memory and executing the instructions, or it can be implemented by a chip circuit.

[0062] In addition, the network architecture and business scenarios described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Ordinary technicians in this field can know that with the evolution of network architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0063] Figure 2 The flowchart of the training method of the repair shop recommendation model for the insurance accident car provided in this application is as follows, wherein the architecture of the repair shop recommendation model includes a first sub-model and a second sub-model; wherein the first sub-model is used to predict the success rate of recommending a comprehensive repair shop, and the second sub-model is used to predict the success rate of recommending all types of repair shops; wherein the first sub-model includes: a first embedding layer, a first feature concatenation layer, and a first fully connected layer; and the second sub-model includes: a second embedding layer, a second feature concatenation layer, and a second fully connected layer. Figure 2 As shown, the method includes:

[0064] S201: extracting a preset number of historical recommended repair data from the repair system.

[0065] Specifically, the accident area of ​​the vehicle is obtained, and the repair shops that can undertake the repair are determined in the area according to the accident area of ​​the vehicle, and the corresponding historical recommended repair data is obtained according to the repair shops that can undertake the repair.

[0066] In this embodiment, the historical recommended repair data includes but is not limited to time, region, vehicle brand, vehicle model, vehicle age, user preference, user's historical damage assessment, geographic location, distance to the repair shop, repair shop category and damage assessment premium ratio.

[0067] S202: Extract features from each piece of historical recommended repair data to obtain first training sample data and second training sample data.

[0068] The first training sample data includes the accident vehicle feature domain, the comprehensive repair shop feature domain, the vehicle-shop cross feature domain and the labeling success rate of the recommended comprehensive repair shop.

[0069] The second training sample data includes the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains and the labeling success rate of all types of repair shops.

[0070] Specifically, each piece of historical recommended repair data is cleaned to remove outliers in the data, and relevant features in the historical recommended repair data are selected for feature extraction to obtain training sample data.

[0071] In this embodiment, the sample training data includes but is not limited to the share of repair shops, brand information, number and amount of repairs, repairs, damage estimates, and damage assessments in different historical periods, the damage assessment amount of new and old cars, signed insurance premiums, brand and model claims ratios, accident vehicle models, types, years, prices, energy labels, usage characteristics, customer preferences for repair shops, historical claims, repair information, maintenance information, satisfaction information, and historical reasons for failed repair recommendations.

[0072] In this embodiment, all types of repair shops include comprehensive repair shops and 4S shop repair shops.

[0073] S203: Input the accident vehicle feature domain, the comprehensive repair shop feature domain, and the vehicle-shop cross feature domain in each first training sample data into the first sub-model, and output the predicted success rate of the comprehensive repair shop after being processed by the first embedding layer, the first feature concatenation layer, and the first fully connected layer in sequence.

[0074] In this embodiment, the first embedding layer is a first embedding feature embedding layer.

[0075] In this embodiment, the features of the sub-model input are based on the historical data accumulated by the accident vehicle repair flow control system, and are input into the first sub-model after data processing.

[0076] In this embodiment, the first fully connected layer includes a first input layer, a first hidden layer and a first output layer.

[0077] In this embodiment, the activation function of the first fully connected layer is the Leaky ReLU activation function.

[0078] S204: Input the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains in each second training sample data into the second sub-model, and output the predicted success rate of all types of repair shops after being processed by the second embedding layer, the second feature concatenation layer and the second fully connected layer in sequence.

[0079] In this embodiment, the second embedding layer is a second embedding feature embedding layer.

[0080] In this embodiment, the features of the sub-model input are based on the historical data accumulated by the accident vehicle repair flow control system, and are input into the second sub-model after data processing.

[0081] In this embodiment, the second fully connected layer includes a second input layer, a second hidden layer and a second output layer.

[0082] In this embodiment, the activation function of the second fully connected layer is a Leaky ReLU activation function.

[0083] S205: Adjust model parameters of the first sub-model and the second sub-model according to the prediction success rate of the comprehensive repair shop and the marking success rate of the comprehensive repair shop, as well as the prediction success rate of all types of repair shops and the marking success rate of all types of repair shops.

[0084] Specifically, the loss functions of the first sub-model and the second sub-model are obtained. If the calculated value of the loss function exceeds a preset range or the predicted success rate of the comprehensive repair shop and the predicted success rate of all types of repair shops exceed the error range, the weight of the corresponding first sub-model or the second sub-model is adjusted according to the value outside the range.

[0085] S206: When the prediction accuracy of the first sub-model and the second sub-model meets the preset standard, stop training to obtain a trained repair shop recommendation model.

[0086] Specifically, the trained repair shop recommendation model is converted into a model in ONNX format to make real-time predictions and recommendations on the repair shops of accident vehicles.

[0087] In this embodiment, the trained repair shop recommendation model is a pytorch model.

[0088] In this embodiment, the repair shop recommendation model pushes repair shops suitable for customers in real time based on user feedback information.

[0089] In this embodiment, the repair shop recommendation model is a deep neural network ranking model that supports multiple business objective tasks.

[0090] It can be seen from the above embodiments that by creating a repair shop recommendation model, two sub-models are set in the repair shop recommendation model, and feature extraction is performed on historical repair recommendation data to obtain first training sample data and second training sample data. The training sample data are input into the two sub-models, and the success rate of recommending comprehensive repair shops is predicted by the first sub-model, and the success rate of recommending all types of repair shops is predicted by the second sub-model. The parameters of the first sub-model and the second sub-model are adjusted to obtain a trained repair shop recommendation model. Compared with the prior art, repair shops can be pushed to insurance customers more accurately, thereby improving customer satisfaction and reducing claims costs.

[0091] In one embodiment of the present application, the first embedding layer of the first sub-model and the second embedding layer of the second sub-model are shared embedding layers; the first embedding layer of the first sub-model is used to input first training sample data and second training sample data; the second embedding layer of the second sub-model is used to input first training sample data and second training sample data.

[0092] In this embodiment, the first embedding layer of the first sub-model and the second embedding layer of the second sub-model share a feature embedding vector dictionary, and through transfer learning, the second sub-model learns from the entire sample space according to the shared feature embedding mechanism.

[0093] Figure 3 Schematic diagram of the recommended model for the repair shop provided for this application.

[0094] like Figure 3 As shown in the figure, the accident vehicle feature domain, repair shop feature domain and vehicle-shop cross feature domain in the training samples are input into two sub-models respectively, the embedding layers of the two sub-models are shared, the outputs of the embedding layers are feature concatenated, and the prediction success rate of the comprehensive repair shop and the prediction success rate of all types of repair shops are output through the fully connected layer, and the auxiliary target success rate is introduced to train the first sub-model.

[0095] It can be seen from the above embodiments that by sharing features of the embedding layers of the two sub-models and through transfer learning, the second sub-model learns from the entire sample space according to the embedding mechanism, thereby reducing the sparsity problem of the second sub-model task training samples.

[0096] In one embodiment of the present application, before step S205, the following steps are included:

[0097] S301: Create an auxiliary target success rate according to the predicted success rate of the comprehensive repair shop and the predicted success rates of all types of repair shops.

[0098] In this embodiment, the predicted success rate of the comprehensive repair shop is recorded as p2, the predicted success rate of all types of repair shops is recorded as p1, and the predicted success rate of the comprehensive repair shop and the predicted success rate of all types of repair shops are multiplied to obtain the auxiliary target success rate p3.

[0099] S302: Train the second sub-model according to the auxiliary target success rate, and calculate and generate a loss function value according to the predicted success rate of the comprehensive repair shop and the predicted success rate of all types of repair shops.

[0100] In this embodiment, the formula for calculating and generating the loss function is:

[0101]

[0102] In the formula, x i Represents the input features; y i Represents the output prediction result; represents a parameter set of the second sub-model; represents the parameter set of the first sub-model; p3 represents the auxiliary target success rate.

[0103] S303: Adjust the second sub-model according to the loss function value to generate a trained second sub-model.

[0104] Specifically, according to the loss function values ​​of p1 and p3, the model is trained, and the model parameters are optimized according to the back propagation algorithm to minimize the sum of the cross entropy loss functions of p1 and p3.

[0105] In this embodiment, p1 and p3 are in the full sample space of the accident car repair shop recommendation model, p2 is used as an intermediate variable, and the supervised information of p1 and p3 is used to train the neural network to implicitly learn p2.

[0106] In this embodiment, the feature characterization parameters of the p2 neural network are shared with the feature characterization parameters of the p1 neural network. Since the training samples of the p1 neural network are relatively abundant, the problem of sparse sample space of p2 is solved by learning the feature characterization parameter transfer.

[0107] In this embodiment, the repair shop recommendation model is designed into two sub-models, p1 and p3 are trained separately, the embedding layer of the input features is shared, the parameters of the sub-models are learned separately, and p2 is used as an auxiliary intermediate variable of the neural network of the repair shop recommendation model.

[0108] It can be seen from the above embodiments that by introducing an auxiliary target success rate into the two sub-models, the implicit learning target p2 of the neural network is trained using the supervisory information of p1 and p3, p2 is used as an intermediate variable, p1 and p3 are predicted in the entire sample space, and p2 is also predicted in the entire space through transfer learning, thereby eliminating the deviation between sample training and model prediction.

[0109] In one embodiment of the present application, the repair shop push repair model is:

[0110] argmax(p1,p2)=f(X1,X2,…,X n )

[0111] Where p1 represents the prediction success rate of all types of repair shops; p2 represents the prediction success rate of comprehensive repair shops; f() represents the linear relationship between p1 and p2; X n Indicates the factors that affect the success rate of the repair.

[0112] It can be seen from the above embodiments that by dividing the deep neural network model of the repair shop recommendation model into a first sub-model and a second sub-model, modeling and training are performed separately to learn the linear relationship between the repair success rate and multiple factors affecting the success rate, the success rate of all types of repair shops and the comprehensive repair shop success rate can be predicted.

[0113] Figure 4A structural diagram of a training device for a repair shop recommendation model for insurance accident vehicles provided in this application, such as Figure 4 As shown, the training device 40 for the repair shop recommendation model for insurance accident vehicles provided in this embodiment includes: a first extraction module 401, a second extraction module 402, a first input module 403, a second input module 404, an adjustment module 405 and a stop module 406.

[0114] The first extraction module 401 is used to extract a preset number of historical recommended repair data from the repair system.

[0115] The second extraction module 402 is used to extract features from each piece of historical recommended repair data to obtain first training sample data and second training sample data.

[0116] The first training sample data includes the accident vehicle feature domain, the comprehensive repair shop feature domain, the vehicle-shop cross feature domain and the labeling success rate of the recommended comprehensive repair shop.

[0117] The second training sample data includes the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains and the labeling success rate of all types of repair shops.

[0118] The first input module 403 is used to input the accident vehicle feature domain, the comprehensive repair shop feature domain and the vehicle-shop cross feature domain in each first training sample data into the first sub-model, and output the predicted success rate of the comprehensive repair shop after being processed by the first embedding layer, the first feature concatenation layer and the first fully connected layer in sequence.

[0119] The second input module 404 is used to input the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains in each second training sample data into the second sub-model, and output the predicted success rate of all types of repair shops after being processed by the second embedding layer, the second feature concatenation layer and the second fully connected layer in sequence.

[0120] The adjustment module 405 is used to adjust the model parameters of the first sub-model and the second sub-model according to the prediction success rate of the comprehensive repair shop and the marking success rate of the comprehensive repair shop, as well as the prediction success rate of all types of repair shops and the marking success rate of all types of repair shops.

[0121] The stopping module 406 is used to stop the training to obtain the trained repair shop recommendation model when the prediction accuracy of the first sub-model and the second sub-model meets the preset standard.

[0122] In one possible implementation, the first embedding layer of the first sub-model and the second embedding layer of the second sub-model are shared embedding layers; the first embedding layer of the first sub-model is used to input the first training sample data and the second training sample data; the second embedding layer of the second sub-model is used to input the first training sample data and the second training sample data.

[0123] In a possible implementation, the device 40 further includes:

[0124] The creation module 407 is used to create an auxiliary target success rate according to the predicted success rate of the comprehensive repair shop and the predicted success rates of all types of repair shops.

[0125] The training module 408 is used to train the second sub-model according to the auxiliary target success rate, and calculate and generate the loss function value according to the predicted success rate of the comprehensive repair shop and the predicted success rate of all types of repair shops.

[0126] The generating module 409 is used to adjust the second sub-model according to the loss function value to generate a trained second sub-model.

[0127] In a possible implementation, the formula for calculating and generating the loss function in the training module 408 is:

[0128]

[0129] In the formula, x i Represents the input features; y i Represents the output prediction result; represents a parameter set of the second sub-model; represents the parameter set of the first sub-model; p3 represents the auxiliary target success rate.

[0130] In a possible implementation, the repair shop promotes the repair model as follows:

[0131] argmax(p1,p2)=f(X1,X2,…,X n )

[0132] Where p1 represents the prediction success rate of all types of repair shops; p2 represents the prediction success rate of comprehensive repair shops; f() represents the linear relationship between p1 and p2; X n Indicates the factors that affect the success rate of the repair.

[0133] The training device for the repair shop recommendation model provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be described in detail here.

[0134] Figure 5This is a schematic diagram of the structure of the training device for the repair shop recommendation model for insurance accident vehicles provided in this application. Figure 5 As shown, the training device 50 for the repair shop recommendation model for insurance accident vehicles provided in this embodiment includes: at least one processor 501 and a memory 502. Optionally, the device 50 also includes a communication component 503. The processor 501, the memory 502 and the communication component 503 are connected via a bus 504.

[0135] In a specific implementation process, at least one processor 501 executes the computer-executable instructions stored in the memory 502, so that at least one processor 501 executes the above-mentioned training method for the repair shop recommendation model for insurance accident vehicles.

[0136] The specific implementation process of the processor 501 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0137] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0138] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-volatile Memory, NVM), such as at least one disk storage.

[0139] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application is not limited to only one bus or one type of bus.

[0140] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned training method for the repair shop recommendation model for insurance accident vehicles.

[0141] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above-mentioned training method for the repair shop recommendation model for insurance accident vehicles is implemented.

[0142] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0143] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0144] The division of units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0145] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0146] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0147] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0148] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0149] Finally, it should be noted that those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art not disclosed by the present invention, are not limited to the precise structure described above and shown in the drawings, and may be modified and changed in various ways without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A training method for a repair shop recommendation model for insurance accident vehicles, characterized in that: The architecture of the repair shop recommendation model includes a first sub-model and a second sub-model; wherein the first sub-model is used to predict the success rate of recommending a comprehensive repair shop, and the second sub-model is used to predict the success rate of recommending all types of repair shops; The first sub-model includes: a first embedding layer, a first feature concatenation layer and a first fully connected layer; the second sub-model includes: a second embedding layer, a second feature concatenation layer and a second fully connected layer; The method comprises: Extract a preset number of historical recommended repair data from the repair system; Perform feature extraction on each piece of historical recommended repair data to obtain first training sample data and second training sample data; The first training sample data includes the accident vehicle feature domain, the comprehensive repair shop feature domain, the vehicle-shop cross feature domain and the labeling success rate of the recommended comprehensive repair shop; The second training sample data includes the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains and the labeling success rate of all types of repair shops; Input the accident vehicle feature domain, the comprehensive repair shop feature domain, and the vehicle-shop cross feature domain in each first training sample data into the first sub-model, and output the predicted success rate of the comprehensive repair shop after being processed by the first embedding layer, the first feature concatenation layer, and the first fully connected layer in sequence; Input the accident vehicle feature domain, various types of repair shop feature domains, and vehicle-shop cross feature domains in each second training sample data into the second sub-model, and output the predicted success rate of all types of repair shops after being processed by the second embedding layer, the second feature concatenation layer, and the second fully connected layer in sequence; Adjusting the model parameters of the first sub-model and the second sub-model according to the prediction success rate of the comprehensive repair shop and the marking success rate of the comprehensive repair shop, as well as the prediction success rate of all types of repair shops and the marking success rate of all types of repair shops; When the prediction accuracy of the first sub-model and the second sub-model meets the preset standard, the training is stopped to obtain a trained repair shop recommendation model.

2. The method according to claim 1, characterized in that The first embedding layer of the first sub-model and the second embedding layer of the second sub-model are shared embedding layers; The first embedding layer of the first sub-model is used to input the first training sample data and the second training sample data; The second embedding layer of the second sub-model is used to input the first training sample data and the second training sample data.

3. The method according to claim 1, characterized in that Before adjusting the model parameters of the first sub-model and the second sub-model according to the prediction success rate of the comprehensive repair shop and the marking success rate of the comprehensive repair shop, and the prediction success rate of all types of repair shops and the marking success rate of all types of repair shops, the method further includes: Creating an auxiliary target success rate according to the predicted success rate of the comprehensive repair shop and the predicted success rates of all types of repair shops; The second sub-model is trained according to the auxiliary target success rate, and a loss function value is calculated and generated according to the predicted success rate of the comprehensive repair shop and the predicted success rate of all types of repair shops; The second sub-model is adjusted according to the loss function value to generate a trained second sub-model.

4. The method according to claim 3, characterized in that The formula for calculating and generating the loss function is: In the formula, x i Represents the input features; y i Represents the predicted result of the output; θ p1 represents the parameter set of the second sub-model; θ p3 represents the parameter set of the first sub-model; p3 represents the auxiliary target success rate.

5. The method according to any one of claims 1 to 4, characterized in that: The recommended model of the repair shop is: argmax(p1,p2)=f(X1,X2,…,X n ) Where p1 represents the prediction success rate of all types of repair shops; p2 represents the prediction success rate of comprehensive repair shops; f() represents the linear relationship between p1 and p2; X n Indicates the factors that affect the success rate of the repair.

6. A training device for a repair shop recommendation model for insurance accident vehicles, characterized in that: The architecture of the repair shop recommendation model includes a first sub-model and a second sub-model; wherein the first sub-model is used to predict the success rate of recommending a comprehensive repair shop, and the second sub-model is used to predict the success rate of recommending all types of repair shops; The first sub-model includes: a first embedding layer, a first feature concatenation layer and a first fully connected layer; the second sub-model includes: a second embedding layer, a second feature concatenation layer and a second fully connected layer; The device comprises: The first extraction module is used to extract a preset number of historical recommended repair data from the repair system; A second extraction module is used to extract features from each piece of historical recommended repair data to obtain first training sample data and second training sample data; The first training sample data includes the accident vehicle feature domain, the comprehensive repair shop feature domain, the vehicle-shop cross feature domain and the labeling success rate of the recommended comprehensive repair shop; The second training sample data includes the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains and the labeling success rate of all types of repair shops; A first input module is used to input the accident vehicle feature domain, the comprehensive repair shop feature domain and the vehicle-shop cross feature domain in each first training sample data into the first sub-model, and output the predicted success rate of the comprehensive repair shop after being processed by the first embedding layer, the first feature concatenation layer and the first fully connected layer in sequence; A second input module is used to input the accident vehicle feature domain, various types of repair shop feature domains and vehicle-shop cross feature domains in each second training sample data into the second sub-model, and output the predicted success rate of all types of repair shops after being processed by the second embedding layer, the second feature concatenation layer and the second fully connected layer in sequence; an adjustment module, configured to adjust the model parameters of the first sub-model and the second sub-model according to the prediction success rate of the comprehensive repair shop and the marking success rate of the comprehensive repair shop, and the prediction success rate of all types of repair shops and the marking success rate of all types of repair shops; The stopping module is used to stop training to obtain a trained repair shop recommendation model when the prediction accuracy of the first sub-model and the second sub-model meets a preset standard.

7. The device according to claim 6, characterized in that The first embedding layer of the first sub-model and the second embedding layer of the second sub-model are shared embedding layers; The first embedding layer of the first sub-model is used to input the first training sample data and the second training sample data; The second embedding layer of the second sub-model is used to input the first training sample data and the second training sample data.

8. A training device for a repair shop recommendation model for insurance accident vehicles, characterized in that: include: at least one processor and memory; The memory stores computer-executable instructions; The at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor performs the training method for a repair shop recommendation model for insurance accident vehicles according to any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the training method for a repair shop recommendation model for insurance accident vehicles as described in any one of claims 1 to 5.

10. A computer program product, characterized in that The invention comprises a computer program, which, when executed by a processor, implements the training method of a repair shop recommendation model for insurance accident vehicles as described in any one of claims 1 to 5.

Images