Service processing model training method and apparatus

Abstract

Embodiments of the present specification provide a service processing model training method and device, wherein the method comprises: determining a first training sample according to service data under a first service type of a target service; training an initial feature processing network in a first processing model under the first service type based on the first training sample to obtain a first feature processing network; performing fusion processing on the first feature processing network and a second feature processing network in a second processing model under a second service type to obtain a feature processing network; and performing model training on a service processing model constructed by a baseline processing network and the feature processing network based on the first training sample and a second training sample under the second service type to obtain the service processing model.

Description

Technical Field

[0001] This document relates to the field of data processing technology, and in particular to a service processing model training method and apparatus. Background Technology

[0002] With the development of network and communication technologies, information networks have become an important part of life. More and more users are participating in various services provided by service providers online. In the process of users participating in different services, the data processed by the service also varies depending on the service.

[0003] With the development of information networks, users are participating in services more and more frequently through the network. More and more service providers are offering different types of services to users. In the process of users participating in different types of services, how to efficiently process data of different types and how to provide users with more efficient and effective services are the key concerns of both users and service providers. Summary of the Invention

[0004] This specification provides one or more embodiments of a service processing model training method. The service processing model training method includes: determining a first training sample based on service data under a first service type of a target service; training an initial feature processing network in a first processing model under the first service type based on the first training sample to obtain a first feature processing network; the first processing model includes a baseline processing network and the initial feature processing network; fusing the first feature processing network with a second feature processing network in a second processing model under a second service type to obtain a feature processing network; and training a service processing model constructed from the baseline processing network and the feature processing network based on the first training sample and the second training sample under the second service type to obtain a service processing model.

[0005] This specification provides one or more embodiments of a service processing method, comprising: acquiring data to be processed for a target service; the data to be processed includes data to be processed under a first service type or a second service type of the target service. The data to be processed is input into a service processing model for service processing to obtain a service processing result. The service processing model is obtained by training a service processing model constructed from a baseline processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type; the feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type.

[0006] This specification provides one or more embodiments of a service processing model training apparatus, comprising: a sample determination module configured to determine a first training sample based on service data under a first service type of a target service; a network training module configured to train an initial feature processing network in a first processing model under the first service type based on the first training sample, to obtain a first feature processing network; the first processing model includes a baseline processing network and the initial feature processing network; a fusion processing module configured to fuse the first feature processing network with a second feature processing network in a second processing model under a second service type, to obtain a feature processing network; and a model training module configured to train a service processing model constructed from the baseline processing network and the feature processing network based on the first training sample and the second training sample under the second service type, to obtain a service processing model.

[0007] This specification provides one or more embodiments of a service processing apparatus, including: a data acquisition module configured to acquire data to be processed for a target service; the data to be processed includes data to be processed under a first service type or a second service type of the target service. The service processing module is configured to input the data to be processed into a service processing model for service processing to obtain a service processing result. The service processing model is obtained by training a service processing model constructed from a baseline processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type; the feature processing network is obtained after incorporating the first feature processing network under the first service type and the second feature processing network under the second service type.

[0008] This specification provides one or more embodiments of a service processing model training apparatus, comprising: a processor; and a memory configured to store computer-executable instructions, which, when executed, cause the processor to: determine a first training sample based on service data under a first service type of a target service; train an initial feature processing network in a first processing model under the first service type based on the first training sample to obtain a first feature processing network; the first processing model includes a baseline processing network and the initial feature processing network; fuse the first feature processing network with a second feature processing network in a second processing model under a second service type to obtain a feature processing network; and train a service processing model constructed from the baseline processing network and the feature processing network based on the first training sample and the second training sample under the second service type to obtain a service processing model.

[0009] This specification provides one or more embodiments of a service processing device, including: a processor; and a memory configured to store computer-executable instructions, which, when executed, cause the processor to: acquire pending data of a target service; the pending data includes pending data under a first service type or a second service type of the target service; input the pending data into a service processing model for service processing to obtain a service processing result. The service processing model is obtained by training a service processing model constructed from a baseline processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type; the feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type.

[0010] This specification provides one or more embodiments of a storage medium for storing computer-executable instructions, which, when executed by a processor, implement the following process: determining a first training sample based on service data under a first service type of a target service; training an initial feature processing network in a first processing model under the first service type based on the first training sample to obtain a first feature processing network; the first processing model includes a baseline processing network and the initial feature processing network; fusing the first feature processing network with a second feature processing network in a second processing model under a second service type to obtain a feature processing network; and training a service processing model constructed from the baseline processing network and the feature processing network based on the first training sample and the second training sample under the second service type to obtain a service processing model.

[0011] This specification provides one or more embodiments of another storage medium for storing computer-executable instructions, which, when executed by a processor, implement the following process: acquiring pending data for a target service; the pending data includes pending data under a first service type or a second service type of the target service. The pending data is then input into a service processing model for service processing to obtain a service processing result. The service processing model is obtained by training a service processing model constructed from a baseline processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type; the feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in one or more embodiments of this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A schematic diagram illustrating an implementation environment provided for one or more embodiments of this specification;

[0014] Figure 2 A flowchart illustrating a service processing model training method provided in one or more embodiments of this specification;

[0015] Figure 3 A schematic diagram of a network before and after fusion processing is provided for one or more embodiments of this specification;

[0016] Figure 4 A flowchart illustrating a service processing model training method for a risk identification model training scenario, provided in one or more embodiments of this specification;

[0017] Figure 5 This specification provides a flowchart of a service processing model training method for a transaction detection model training scenario, which is provided in one or more embodiments.

[0018] Figure 6 A service processing method flowchart provided for one or more embodiments of this specification;

[0019] Figure 7 A schematic diagram of an embodiment of a service processing model training device provided in one or more embodiments of this specification;

[0020] Figure 8 A schematic diagram of an embodiment of a service processing apparatus provided in one or more embodiments of this specification;

[0021] Figure 9 A schematic diagram of the structure of a service processing model training device provided in one or more embodiments of this specification;

[0022] Figure 10 This is a schematic diagram of the structure of a service processing device provided for one or more embodiments of this specification. Detailed Implementation

[0023] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0024] The service processing model training method provided in one or more embodiments of this specification is applicable to the implementation environment of training a service processing model, such as... Figure 1 As shown, the implementation environment includes at least server 101, benchmark processing network 102 and service processing model 103. The benchmark processing network 102 and service processing model can be deployed on server 101, or corresponding servers can be allocated separately for the benchmark processing network 102 and service processing model 103 so that the benchmark processing network 102 and service processing model 103 are deployed on their respective allocated servers.

[0025] In addition, the implementation environment may also include a user terminal 104, through which service data can be determined and labeled to obtain training samples.

[0026] The server 101 can be one or more servers, a server cluster consisting of several servers, or a cloud server of a cloud computing platform, used to train and call the service processing model 103. In addition, the server 101 can also be used to generate and manage service data of the target service. The user terminal 104 can be a smartphone, tablet computer, e-book reader, wearable device, or device that interacts with information based on AR (Augmented Reality) / VR (Virtual Reality), etc. It can also have an application or browser installed, and can determine and label service data by interacting with the user through the application or browser, or by interacting with the user through subroutines in the application.

[0027] In this implementation environment, server 101 trains the initial feature processing network in the first processing model composed of the baseline processing network and the feature processing network under the first service type based on the first training samples under the first service type, to obtain the first feature processing network under the first service type of the target service. The first feature processing network and the second feature processing network under the second service type are merged to obtain the feature processing network. Based on the first training samples under the first service type and the second training samples under the second service type, the service processing model constructed by the baseline processing network and the feature processing network is trained to obtain the trained service processing model.

[0028] This specification provides one or more embodiments of a service processing model training method, as follows:

[0029] This specification provides a service processing model training method. After training the model for each service type of the target service to obtain the processing model for each service type, the feature processing networks in the processing models of each service type are fused to obtain a feature processing network. Based on the baseline processing network and the feature processing network of the target service, a service processing model for the target service is constructed, so that a service can achieve service processing for different service types by deploying only one model. Furthermore, the constructed service processing model is further trained using training samples under each service type to improve the effectiveness and accuracy of the service processing model in processing each service type.

[0030] Reference Figure 2 The service processing model training method provided in this embodiment specifically includes steps S202 to S208.

[0031] Step S202: Determine the first training sample based on the service data under the first service type of the target service.

[0032] In this embodiment, the target service includes services provided to individual users or groups such as merchants; for example, risk identification services and transaction services. The data involved in the same service may differ depending on whether the user or merchant participates, and the service processing methods may also differ. Therefore, the type of target service differs depending on whether the user or merchant participates.

[0033] For example, in the risk identification service, risk identification is performed on the user's username, identity credentials, and other information, and on the merchant's proof of merchant status. Here, different service types are used for different processing objects. As another example, in the transaction service, when a user transfers funds to another user, the relationship between the two users is verified; when a user issues rights to another user, the permissions of both the user and the other user for those rights are verified. Here, different service types are used for different service forms. In this embodiment, the first service type and the second service type represent different service types under the target service. The service type of the target service can be divided based on the objects participating in the target service, or based on the different services provided by the target service. Furthermore, it can be divided based on other factors, and the specific configuration can be tailored to the actual scenario. This embodiment does not impose any limitations on this.

[0034] In this embodiment, the data obtained or generated under different service types of the target service is the service data under that service type.

[0035] In practice, in order to train a model that can process the data to be processed under the first service type, the first training sample is determined based on the service data under the first service type of the target service; for example, user data under the user risk identification type in the risk identification service is used as sample data, and the risk category corresponding to each user data is used as sample label to obtain the training sample under the user risk identification type in the risk identification service.

[0036] Step S204: Based on the first training samples, train the initial feature processing network in the first processing model under the first service type to obtain the first feature processing network.

[0037] In the target service, the data processed may differ due to different service types, but the service itself remains consistent. For example, whether it's risk identification for users or merchants, it's still risk identification; whether it's money transfer or rights distribution, it's essentially resource transfer. Therefore, a baseline processing network can be configured for the target service to handle its service processing. Since the baseline processing network is, to some extent, a large model, training it for each different service type would lead to resource waste. Furthermore, if only a large model is trained, it might suffer from forgetting issues after learning new tasks, resulting in poor performance in previous service types. Therefore, based on the baseline processing network for the target service, feature processing networks are trained for each service type.

[0038] The baseline processing network in this embodiment refers to a relatively general large model with a lot of basic knowledge, that is, a trained large model with certain processing capabilities; it can be an open-source large model, a large model developed for the target service, or a large model further developed based on an open-source large model.

[0039] In this embodiment, the initial feature processing network refers to a network that performs feature processing without prior training; the first feature processing network refers to a network that performs feature processing under the first service type, obtained by training the initial feature processing network based on the first training samples under the first service type.

[0040] In this embodiment, the network performing feature processing includes a network for feature encoding, such as LoRA (Low-Rank Adaptation of Large Language Models). The data to be processed is input into the feature processing network for feature processing, and then input into the baseline processing network. This allows the output of the baseline processing network to more closely match the processing requirements of the corresponding service type. It should be noted that the baseline processing network can include multiple layers, and the feature processing network can also include multiple layers. The feature processing network and the baseline processing network can be deployed intermittently. For example, if the baseline processing network includes 10 layers, the feature processing network can be inserted after the 1st, 4th, and 7th layers of the baseline processing network, respectively. Alternatively, the feature processing network can be inserted after each layer of the baseline processing network. Specifically, the deployment relationship between the baseline processing network and the feature processing network can be configured according to actual needs, and this embodiment does not impose any limitations. However, it should be noted that under the target service, the deployment method of the baseline processing network and the initial feature processing network is consistent across different service type processing models.

[0041] In this embodiment, to ensure the processing effect of the data to be processed under the first service type of the target service, the initial feature processing network in the first processing model under the first service type is trained based on the first training samples to obtain the first feature processing network. Optionally, the first processing model includes a baseline processing network and the initial feature processing network.

[0042] Specifically, during the training of the first processing model for the first service type, only the initial feature processing network within the first processing model is trained. This avoids training the baseline processing network, which could reduce its processing performance on data from other service types. It should be noted that since the baseline processing network is primarily responsible for service processing, the training loss is still calculated based on the loss function corresponding to the baseline processing network during the training of the initial feature processing network.

[0043] In the specific execution process, in order to avoid storing processing models for each service type of the target service, which would result in too many models being stored for the target service and occupying a large amount of storage space, the initial feature processing network in the processing model under each service model is trained to obtain the feature processing network for each service type. Then, the feature processing networks of each service type are fused to obtain the feature processing network for the target service.

[0044] Based on this, in the first optional implementation provided in this embodiment, during the training of the initial feature processing network in the first processing model under the first service type based on the first training samples, the following operations are performed:

[0045] The sample data in the first training sample is input into the first processing model for service processing to obtain the first service processing result;

[0046] Calculate the training loss based on the first service processing result and the sample labels in the first training sample;

[0047] Based on the training loss, the parameters of the initial feature processing network in the first processing model are adjusted.

[0048] Specifically, the sample data in the first training sample is input into the first processing model for service processing to obtain the first service processing result. The first service processing result and the sample labels in the first training sample are input into the loss function for loss calculation to obtain the training loss. Based on the training loss, the parameters of the initial feature processing network in the first processing model are adjusted until the first processing model converges to obtain the first feature processing network.

[0049] In other words, a combination of a baseline processing network and an initial feature processing network is used. During parameter adjustment, the parameters of the baseline processing network are frozen, and only the parameters of the initial feature processing network are adjusted. If the target service only includes the first service type, after training the initial feature processing network in the first processing model under the first service type, the first processing model under the first service type can be used as the service processing model for the target service.

[0050] In the second optional implementation provided in this embodiment, the process of training the initial feature processing network in the first processing model under the first service type based on the first training samples to obtain the first feature processing network is carried out in the following manner:

[0051] Based on the first training sample, the initial feature output network in the initial feature processing network of the first processing model is trained to obtain the first feature output network;

[0052] The first feature processing network is constructed based on the feature transformation network and the first feature output network in the initial feature processing network.

[0053] Specifically, the initial feature processing network can consist of a feature transformation network and a feature output network. During training, only the parameters of the feature output network are adjusted, and the first feature output network is constructed based on the obtained first feature output network and the feature transformation network in the initial feature processing network.

[0054] Optionally, based on the first training samples, the initial feature output network in the initial feature processing network of the first processing model is trained to obtain the first feature output network. During this process, sample data from the first training samples are input into the first processing model for service processing to obtain the first service processing result. Based on the first service processing result and the sample labels in the first training samples, the training loss is calculated. According to the training loss, the parameters of the initial feature output network in the initial feature processing network of the first processing model are adjusted until the model converges to obtain the first feature output network.

[0055] In other words, a mode combining a baseline processing network, a feature transformation network, and an initial feature output network is used. During the parameter adjustment process, the parameters of the baseline processing network and the feature transformation network are frozen, and only the parameters of the initial feature output network are adjusted.

[0056] In addition to the two methods mentioned above for training the initial feature processing network, a combination of the two methods can also be used. For example, during the training of the initial feature network in the first processing model, the following operations can be performed:

[0057] The sample data in the first training sample is input into the first processing model for service processing to obtain the first service processing result;

[0058] Calculate the training loss based on the first service processing result and the sample labels in the first training sample;

[0059] Based on the training loss, the parameters of the initial feature processing network in the first processing model are adjusted.

[0060] Optionally, the initial feature processing network consists of an initial feature transformation network and an initial feature output network. The parameters of the initial feature processing network in the first processing model are adjusted until the model converges, and the resulting first feature processing network consists of a first feature transformation network and a first feature output network.

[0061] Step S206: The first feature processing network and the second feature processing network in the second processing model under the second service type are fused to obtain the feature processing network.

[0062] The feature processing network includes a network for performing feature processing on the target service.

[0063] Similar to the training process of the first feature processing network of the first processing model under the first service type described above, in an optional implementation provided in this embodiment, the second feature processing network is obtained in the following manner:

[0064] Based on the second training samples, the initial feature processing network in the second processing model is trained to obtain the second feature processing network;

[0065] The second processing model includes the baseline processing network and the initial feature processing network.

[0066] In practice, in order to save storage space occupied by the target service, after obtaining the first feature processing network under the first service type and the second feature processing network under the second service type of the target service, the first feature processing network and the second feature processing network are fused to obtain the feature processing network of the target service.

[0067] Corresponding to the two training methods for the initial feature processing network mentioned above, the fusion processing in this embodiment includes two types: parameter fusion and network merging. The following is a detailed explanation of the two fusion processing methods provided in this embodiment.

[0068] (1) Parameter fusion

[0069] Corresponding to the training method of the first initial feature processing network provided above, in an optional implementation of this embodiment, the fusion processing of the first feature processing network and the second feature processing network is achieved in the following manner:

[0070] Read the first network parameters of the first feature processing network and the second network parameters of the second feature processing network;

[0071] The first network parameters and the second network parameters are fused to obtain the target network parameters;

[0072] The feature processing network is constructed based on the target network parameters.

[0073] Specifically, the target network parameters are obtained by fusing the first network parameters of the first feature processing network and the second network parameters of the second feature processing network, and the feature processing network whose network parameters are the target network parameters is used as the feature processing network for the target service.

[0074] In other words, the feature processing networks of each service type of the target service are fused together to obtain the feature processing network of the target service; specifically, the average value of the network parameters of the feature processing networks of each service type is calculated and used as the target network parameters.

[0075] like Figure 3 As shown in (a), the first feature processing network and the second feature processing network are fused to obtain a single feature processing network. The structure of the feature processing network is the same as that of the first and second feature processing networks. For example, if the first feature processing network includes a first feature transformation network and a first feature output network, and the second feature processing network includes a second feature transformation network and a second feature output network, then the feature processing network obtained by the fusion process includes a feature transformation network and a feature output network.

[0076] (2) Network Merging

[0077] Corresponding to the training method of the second initial feature processing network provided above, in an optional implementation of this embodiment, the fusion processing of the first feature processing network and the second feature processing network is achieved in the following manner:

[0078] The first feature output network and the second feature processing network are merged to obtain a merged output network.

[0079] The feature processing network is constructed based on the feature transformation network and the merged output network.

[0080] Specifically, since there are two feature output networks, in order to achieve network merging, the first feature output network and the second feature output network are merged to obtain a merged output network. Then, a feature processing network is constructed based on the feature transformation network and the merged output network to obtain a feature processing network that includes the feature transformation network, the first feature processing network, and the second feature processing network.

[0081] like Figure 3 As shown in (b), after the first and second feature processing networks are fused, the resulting feature processing network includes one feature transformation network and two feature output networks. The structure of this feature processing network differs from that of the first and second feature processing networks. For example, if the first feature processing network includes a first feature transformation network and a first feature output network, and the second feature processing network includes a second feature transformation network and a second feature output network, then the feature processing network obtained through fusion processing includes one feature transformation network and two feature output networks.

[0082] In addition to performing parameter fusion and network merging separately as described above to achieve fusion processing, parameter fusion and network merging can also be combined to achieve fusion processing of the first feature processing network and the second feature processing network; corresponding to the above-mentioned combined training process of the initial feature processing network, in this embodiment, the fusion processing of the first feature processing network and the second feature processing network can also be performed based on the following method:

[0083] Read the first transformation network parameters of the first feature transformation network in the first feature processing network, and read the second transformation network parameters of the second feature transformation network in the second feature processing network; fuse the first transformation network parameters and the second transformation network parameters to obtain the target transformation network parameters; construct the target feature transformation network based on the target transformation network parameters; and,

[0084] The first feature output network in the first feature processing network and the second feature output network in the second feature processing network are merged to obtain a merged output network;

[0085] The feature processing network is constructed based on the target feature transformation network and the merged output network.

[0086] That is to say, Figure 3 The feature transformation network shown in (b) is the first feature transformation network when the first feature transformation network in the first feature processing network is the same as the second feature transformation network in the second feature processing network; when the first feature transformation network and the second feature transformation network are not the same, it is the feature transformation network after the first feature transformation network and the second feature transformation network are fused.

[0087] Step S208: Based on the first training samples and the second training samples under the second service type, train the service processing model constructed by the benchmark processing network and the feature processing network to obtain the service processing model.

[0088] After fusing the first and second feature processing networks to obtain the feature processing network, a service processing model is constructed based on the benchmark processing network and the feature processing network to achieve service processing for the target service. After obtaining the service processing model, the model is trained using the first training samples and the second training samples under the second service type to obtain the trained service processing model, which is then used to process the data to be processed under the target service, ensuring the service processing effect of the service processing model under each service type.

[0089] In specific implementation, if the obtained feature processing network is a feature processing network after parameter fusion, that is, the feature processing network has the same structure as the first feature processing network and the second feature processing network, then during the model training process of the service processing model, the service processing model is trained using the first training sample and the second training sample, so that the trained service processing model has the ability to process the data to be processed under the first service type and the second service type.

[0090] In one optional implementation of this embodiment, the following operations are performed during the process of training the service processing model constructed by the benchmark processing network and the feature processing network based on the first training samples and the second training samples under the second service type to obtain the service processing model:

[0091] The first training sample is sampled according to a preset ratio to obtain a first sampled sample, and the second training sample is sampled according to the preset ratio to obtain a second sampled sample;

[0092] Based on the first sample and the second sample, the feature processing network in the service processing network constructed by the benchmark processing network and the feature processing network is trained to obtain the trained service processing model.

[0093] In the specific implementation process, in order to avoid the training time being too long and resources being wasted due to the large number of first and second training samples, the first and second training samples are sampled, and the service processing model is trained based on the sampled training samples. In order to make the service processing model trained to perform similarly on the first and second service types, the first and second training samples are sampled according to a preset ratio to obtain the first sampled sample and the second sampled sample. The service processing model is trained based on the first sampled sample and the second sampled sample.

[0094] During the training of the service processing model, the parameters of the baseline processing network remain unchanged, while the parameters of the feature processing network are adjusted.

[0095] After training the service processing model, for the data to be processed under each service type of the target service, the service processing model is used for service processing. In an optional implementation method provided in this embodiment, the following operations are also performed:

[0096] Obtain the pending data of the target service; the pending data includes data under the first service type or the second service type.

[0097] The data to be processed is input into the trained service processing model for service processing to obtain the service processing result.

[0098] In another optional implementation provided in this embodiment, the service processing model constructed by the benchmark processing network and the feature processing network can also be trained based on the first training samples and the second training samples under the second service type to obtain the service processing model:

[0099] Based on the first training samples, the first feature output network in the service processing model to be trained is trained, and based on the second training samples, the second feature output network in the service processing model to be trained is trained to obtain the service processing model.

[0100] Specifically, when the obtained feature processing network is a feature processing network obtained after network merging, that is, when the structure of the feature processing network is different from that of the first feature processing network and the second feature processing network, and the feature processing network includes feature output networks for each service type, during the model training process of the service processing model, the first feature output network in the service processing model is trained using the first training samples, and the second feature output network in the service processing model is trained using the second training samples, so that the trained service processing model outputs the service processing results of each service type through the feature output networks corresponding to each service type.

[0101] In addition to training the first feature output network in the service processing model to be trained based on the first training samples, and the second feature output network in the service processing model to be trained based on the second training samples, the feature transformation network and the first feature output network in the service processing model to be trained can also be trained based on the first training samples, and the feature transformation network and the second feature output network in the service processing model to be trained can also be trained based on the second training samples to obtain the service processing model. The service processing model trained based on the second training samples can be one that has already been trained using the first training samples. It should also be noted that, during the model training process based on the first and second training samples, regardless of the implementation method described above, the first training samples can be sampled according to a preset ratio to obtain the first sampled samples, and the second training samples can be sampled according to the preset ratio to obtain the second sampled samples. The service processing model to be trained is then trained based on the first and second sampled samples.

[0102] In this embodiment, the service processing model to be trained is the service processing model constructed from the baseline processing network and the feature processing network.

[0103] After training a service processing model containing feature output networks for each service type, service processing can be performed on the data to be processed under each service type of the target service based on this model. It should be noted that this service processing model includes a type partitioning network, which partitions the data to be processed into service types. After obtaining the service types of the data to be processed, the model processes the data based on a baseline processing network, a feature transformation network, and the feature output network corresponding to the service type, thus obtaining the service processing result. Since the feature transformation network and the feature output network are deployed in different positions on the baseline processing network, each deployed feature transformation network and feature output network processes the data based on the feature transformation network and the feature output network corresponding to the service type, and then outputs the final network. This embodiment describes the service processing process in detail with the feature processing network deployed before the baseline processing network. In one optional implementation of this embodiment, after obtaining the trained service processing model, the data to be processed for the target service is input into the trained service processing model for service processing to obtain the service processing result.

[0104] The service processing includes:

[0105] The data to be processed is classified by type to obtain the service type of the data to be processed;

[0106] The data to be processed is input into a trained feature transformation network to perform feature transformation, thereby obtaining transformed features;

[0107] The transformation features are input into the feature output network corresponding to the service type for output calculation to obtain the output features;

[0108] The output features are input into the benchmark processing network for feature processing to obtain the service processing result.

[0109] Specifically, firstly, the type classification network classifies the data to be processed into service types; then, the feature transformation network transforms the data to be processed into transformed features; the feature output network corresponding to the service type calculates the output features of the transformed features; and the baseline processing network processes the output features to obtain the service processing result.

[0110] It should be noted that the data to be processed in this embodiment may be data obtained or generated during the service participation process, such as user information, merchant information, and transaction information.

[0111] In summary, the service processing model training method provided in this embodiment determines training samples for each service type based on service data under each service type of the target service. Based on the training samples under each service type, the initial feature processing network in the processing model under each service type is trained to obtain the feature processing network for each service type. The feature processing networks of each service type are fused to obtain the feature processing network for the target service. A service processing model is constructed based on the baseline processing network and the feature processing network of the target service. In order to enable the obtained service processing model to perform service processing on the data to be processed under each service type simultaneously, the service processing model is trained based on the training samples under each service type to obtain the trained service processing model, which is then used to perform service processing on the data to be processed under each service type of the target service.

[0112] The following description uses the application of a service processing model training method provided in this embodiment in a risk identification model training scenario as an example to further illustrate the service processing model training method provided in this embodiment. (See also...) Figure 4 The service processing model training method applied to risk identification model training scenarios includes the following steps.

[0113] Step S402: Determine user training samples based on user data under the user risk identification type of the risk identification service.

[0114] Step S404: Based on user training samples, train the initial feature processing network in the user risk identification model under the user risk identification type to obtain the user feature processing network.

[0115] Optionally, the user risk identification model includes a risk identification network and an initial feature processing network.

[0116] Step S406: Read the merchant feature processing network in the merchant risk identification model under the merchant risk identification type of the risk identification service.

[0117] Optionally, the merchant risk identification model is trained using a process similar to that described above, prior to the training of the user risk identification model.

[0118] Step S408: The network parameters of the user feature processing network and the network parameters of the merchant feature processing network are fused to obtain the target network parameters.

[0119] Step S410: Construct a feature processing network based on the target network parameters, and construct an initial risk identification model based on the risk identification network and the feature processing network.

[0120] Optionally, the initial risk identification model has the same model structure as the user risk identification model and the merchant risk identification model.

[0121] Step S412: Sample the user training samples to obtain user sample samples, and sample the merchant training samples to obtain merchant sample samples.

[0122] Step S414: Train the feature processing network in the initial risk identification model based on user sample samples and merchant sample samples to obtain the risk identification model.

[0123] Furthermore, after obtaining the risk identification model, if user data to be identified is acquired, the user data to be identified is input into the risk identification model for risk identification to obtain the user risk identification result; if merchant data to be identified is acquired, the merchant data to be identified is input into the risk identification model for risk identification to obtain the merchant risk identification result.

[0124] The following description uses the application of a service processing model training method provided in this embodiment in a transaction detection model training scenario as an example to further illustrate the service processing model training method provided in this embodiment. (See also...) Figure 5 The service processing model training method applied to the transaction detection model training scenario includes the following steps.

[0125] Step S502: Determine the transfer training samples based on the transfer data under the transfer detection type of the transaction detection service.

[0126] Step S504: Based on the transfer training samples, train the initial feature processing network in the transfer detection model under the transfer detection type to obtain the transfer feature processing network.

[0127] Optionally, the transfer detection model includes a transaction detection network and an initial feature processing network.

[0128] Step S506: Read the payment feature processing network in the payment detection model under the payment detection type of the transaction detection service.

[0129] Optionally, the payment detection model is trained using a similar process to that described above before training the transfer detection model.

[0130] Step S508: Merge the payment feature processing network and the transfer feature processing network to obtain a merged processing network.

[0131] Step S510: Construct an initial transaction detection model based on the transaction detection network and the merge processing network.

[0132] Optionally, the merged processing network of the initial transaction detection model includes a transfer feature processing network and a payment feature processing network.

[0133] Step S512: Sample the transfer training samples to obtain transfer sample samples, and sample the payment training samples to obtain payment sample samples.

[0134] Step S514: Train the transfer feature processing network in the initial transaction detection model based on the transfer sampling samples, and train the payment feature processing network in the initial transaction detection model based on the payment sampling samples to obtain the transaction detection model.

[0135] Furthermore, after obtaining the transaction detection model, if the transfer data to be detected is acquired, the transfer data to be detected is output to the transaction detection model for transaction detection, and the detection result of transaction access or non-access is obtained.

[0136] Transaction detection includes:

[0137] The type classification layer in the transaction detection network classifies the transfer data to be detected into types, thus obtaining the transfer detection type.

[0138] Transfer detection is performed based on other layers in the transaction detection network and the transfer feature processing network corresponding to the transfer detection type, and the transfer detection results are obtained.

[0139] One or more embodiments of a service processing method provided in this specification are as follows:

[0140] The service processing method provided in this embodiment is similar to the service processing model training method provided in the above embodiments. When reading this embodiment, please refer to the relevant content of the above embodiments or make adaptive modifications to the relevant content of the above embodiments. This embodiment will not be described in detail here.

[0141] Reference Figure 6 The service processing method provided in this embodiment specifically includes steps S602 to S604.

[0142] Step S602: Obtain the data to be processed for the target service.

[0143] Optionally, the data to be processed includes data to be processed under the first service type or the second service type of the target service.

[0144] Step S604: Input the data to be processed into the service processing model for service processing to obtain the service processing result.

[0145] Optionally, the service processing model is obtained by training a service processing model constructed from a benchmark processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type; the feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type.

[0146] In one optional implementation of this embodiment, the service processing includes:

[0147] The data to be processed is input into a feature transformation network for feature transformation to obtain transformed features;

[0148] The transformed features are input into the feature output network for output calculation to obtain the output features;

[0149] The output features are input into the benchmark processing network for feature processing to obtain the service processing result.

[0150] In another optional implementation provided in this embodiment, the service processing includes:

[0151] The data to be processed is classified by type to obtain the service type of the data to be processed;

[0152] The data to be processed is input into a feature transformation network for feature transformation to obtain transformed features;

[0153] The transformation features are input into the feature output network corresponding to the service type for output calculation to obtain the output features;

[0154] The output features are input into the benchmark processing network for feature processing to obtain the service processing result.

[0155] This specification provides one or more embodiments of a service processing model training device as follows:

[0156] In the above embodiments, a service processing model training method is provided, and correspondingly, a service processing model training device is also provided, which will be described below with reference to the accompanying drawings.

[0157] Reference Figure 7 This illustration shows a schematic diagram of an embodiment of a service processing model training device provided in this embodiment.

[0158] Since the apparatus embodiments correspond to the method embodiments, the descriptions are relatively simple. For relevant parts, please refer to the corresponding descriptions of the method embodiments provided above. The apparatus embodiments described below are merely illustrative.

[0159] This embodiment provides a service processing model training device, including:

[0160] The sample determination module 702 is configured to determine the first training sample based on the service data under the first service type of the target service.

[0161] The network training module 704 is configured to train the initial feature processing network in the first processing model under the first service type based on the first training samples, thereby obtaining the first feature processing network; the first processing model includes a baseline processing network and the initial feature processing network.

[0162] The fusion processing module 706 is configured to fuse the first feature processing network and the second feature processing network in the second processing model under the second service type to obtain a feature processing network.

[0163] The model training module 708 is configured to train the service processing model constructed by the benchmark processing network and the feature processing network based on the first training samples and the second training samples under the second service type, so as to obtain the service processing model.

[0164] This specification provides one or more embodiments of a service processing device as follows:

[0165] In the above embodiments, a service processing method is provided, and correspondingly, a service processing apparatus is also provided, which will be described below with reference to the accompanying drawings.

[0166] Reference Figure 8 This illustration shows a schematic diagram of a service processing device embodiment provided in this embodiment.

[0167] Since the apparatus embodiments correspond to the method embodiments, the descriptions are relatively simple. For relevant parts, please refer to the corresponding descriptions of the method embodiments provided above. The apparatus embodiments described below are merely illustrative.

[0168] This embodiment provides a service processing device, including:

[0169] The data acquisition module 802 is configured to acquire pending data of the target service; the pending data includes pending data under the first service type or the second service type of the target service.

[0170] Service processing module 804 is configured to input the data to be processed into the service processing model for service processing and obtain the service processing result;

[0171] The service processing model is obtained by training a service processing model constructed from a benchmark processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type. The feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type.

[0172] This specification provides one or more embodiments of a service processing model training device as follows:

[0173] Corresponding to the service processing model training method described above, based on the same technical concept, one or more embodiments of this specification also provide a service processing model training device, which is used to execute the service processing model training method provided above. Figure 9 This is a schematic diagram of the structure of a service processing model training device provided in one or more embodiments of this specification.

[0174] This embodiment provides a service processing model training device, comprising:

[0175] like Figure 9 As shown, the service processing model training device can vary significantly due to differences in configuration or performance. It may include one or more processors 901 and memory 902, and the memory 902 may store one or more application programs or data. The memory 902 may be temporary or persistent storage. The application programs stored in the memory 902 may include one or more modules (not shown), each module may include a series of computer-executable instructions from the service processing model training device. Furthermore, the processor 901 may be configured to communicate with the memory 902 and execute the series of computer-executable instructions in the memory 902 on the service processing model training device. The service processing model training device may also include one or more power supplies 903, one or more wired or wireless network interfaces 904, one or more input / output interfaces 905, one or more keyboards 906, etc.

[0176] In one specific embodiment, the service processing model training apparatus includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the service processing model training apparatus, and is configured to be executed by one or more processors. The one or more programs include computer-executable instructions for performing the following:

[0177] The first training sample is determined based on the service data under the first service type of the target service.

[0178] Based on the first training samples, the initial feature processing network in the first processing model under the first service type is trained to obtain the first feature processing network; the first processing model includes a baseline processing network and the initial feature processing network.

[0179] The first feature processing network and the second feature processing network in the second processing model under the second service type are fused together to obtain the feature processing network.

[0180] Based on the first training samples and the second training samples under the second service type, the service processing model constructed by the benchmark processing network and the feature processing network is trained to obtain the service processing model.

[0181] This specification provides one or more embodiments of a service processing device as follows:

[0182] Corresponding to the service processing method described above, based on the same technical concept, one or more embodiments of this specification also provide a service processing device for executing the service processing method provided above. Figure 10 This is a schematic diagram of the structure of a service processing device provided for one or more embodiments of this specification.

[0183] This embodiment provides a service processing device, including:

[0184] like Figure 10 As shown, the service processing device can vary considerably due to differences in configuration or performance. It may include one or more processors 1001 and memory 1002, and the memory 1002 may store one or more application programs or data. The memory 1002 may be temporary or persistent storage. The application programs stored in the memory 1002 may include one or more modules (not shown), each module may include a series of computer-executable instructions from the service processing device. Furthermore, the processor 1001 may be configured to communicate with the memory 1002 and execute the series of computer-executable instructions in the memory 1002 on the service processing device. The service processing device may also include one or more power supplies 1003, one or more wired or wireless network interfaces 1004, one or more input / output interfaces 1005, one or more keyboards 1006, etc.

[0185] In one specific embodiment, the service processing device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for use in the service processing device, and is configured to be executed by one or more processors. The one or more programs include computer-executable instructions for performing the following:

[0186] Obtain pending data for the target service; the pending data includes pending data under the first service type or the second service type of the target service.

[0187] The data to be processed is input into the service processing model for service processing to obtain the service processing result.

[0188] The service processing model is obtained by training a service processing model constructed from a benchmark processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type. The feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type.

[0189] This specification provides one or more embodiments of a storage medium as follows:

[0190] Corresponding to the service processing model training method described above, based on the same technical concept, one or more embodiments of this specification also provide a storage medium.

[0191] The storage medium provided in this embodiment is used to store computer-executable instructions, which, when executed by a processor, implement the following process:

[0192] The first training sample is determined based on the service data under the first service type of the target service.

[0193] Based on the first training samples, the initial feature processing network in the first processing model under the first service type is trained to obtain the first feature processing network; the first processing model includes a baseline processing network and the initial feature processing network.

[0194] The first feature processing network and the second feature processing network in the second processing model under the second service type are fused together to obtain the feature processing network.

[0195] Based on the first training samples and the second training samples under the second service type, the service processing model constructed by the benchmark processing network and the feature processing network is trained to obtain the service processing model.

[0196] It should be noted that the embodiments concerning storage media in this specification and the embodiments concerning service processing model training methods in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can be referred to the implementation of the corresponding methods described above, and the repeated parts will not be described again.

[0197] One or more embodiments of another storage medium provided in this specification are as follows:

[0198] Corresponding to the service processing method described above, and based on the same technical concept, one or more embodiments of this specification also provide a storage medium.

[0199] The storage medium provided in this embodiment is used to store computer-executable instructions, which, when executed by a processor, implement the following process:

[0200] Obtain pending data for the target service; the pending data includes pending data under the first service type or the second service type of the target service.

[0201] The data to be processed is input into the service processing model for service processing to obtain the service processing result.

[0202] The service processing model is obtained by training a service processing model constructed from a benchmark processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type. The feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type.

[0203] It should be noted that the embodiments concerning storage media and the embodiments concerning service processing methods in this specification are based on the same inventive concept. Therefore, the specific implementation of this embodiment can be referred to the implementation of the corresponding method described above, and the repeated parts will not be described again.

[0204] The various embodiments in this specification are described in a progressive manner. For the same or similar parts between the various embodiments, please refer to each other. Each embodiment focuses on describing the differences from other embodiments. For example, the device embodiment, equipment embodiment, and storage medium embodiment are all similar to the method embodiment, so the description is relatively simple. For reading the relevant content of the device embodiment, equipment embodiment, and storage medium embodiment, please refer to the description of the method embodiment.

[0205] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0206] In the 1930s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many improvements to the methodology today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that an improvement to the methodology cannot be implemented using a hardware physical module. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program a digital system themselves to "integrate" it onto a PLD, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0207] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0208] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0209] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, when implementing the embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0210] Those skilled in the art will understand that one or more embodiments of this specification can be provided as a method, system, or computer program product. Therefore, one or more embodiments of this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0211] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0212] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0213] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0214] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0215] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0216] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0217] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0218] One or more embodiments of this specification can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a particular task or implement a particular abstract data type. One or more embodiments of this specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0219] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0220] The above description is merely an embodiment of this document and is not intended to limit the scope of this document. Various modifications and variations can be made to this document by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this document should be included within the scope of the claims of this document.

Claims

1. A service processing model training method, comprising: The first training sample is determined based on the service data under the first service type of the target service, wherein the target service includes risk identification service, the first service type includes user risk identification type, and the first training sample includes user data and risk categories corresponding to each user data. Based on the first training sample, the initial feature processing network in the first processing model under the first service type is trained to obtain the first feature processing network; The first processing model includes a baseline processing network and the initial feature processing network; The first feature processing network and the second feature processing network in the second processing model under the second service type of the target service are fused to obtain a feature processing network. The second feature processing network is obtained by training the initial feature processing network in the second processing model based on the second training samples under the second service type. The first service type and the second service type represent different service types under the target service. The second service type includes merchant risk identification type. The second training samples include merchant data and the risk category corresponding to each merchant data. Based on the first training samples and the second training samples under the second service type, the service processing model constructed by the benchmark processing network and the feature processing network is trained to obtain the service processing model.

2. The method according to claim 1, wherein training the initial feature processing network in the first processing model under the first service type based on the first training samples includes: The sample data in the first training sample is input into the first processing model for service processing to obtain the first service processing result; Calculate the training loss based on the first service processing result and the sample labels in the first training sample; Based on the training loss, the parameters of the initial feature processing network in the first processing model are adjusted.

3. The method according to claim 1, wherein fusing the first feature processing network and the second feature processing network in the second processing model under the second service type to obtain a feature processing network comprises: Read the first network parameters of the first feature processing network and the second network parameters of the second feature processing network; The first network parameters and the second network parameters are fused to obtain the target network parameters; The feature processing network is constructed based on the target network parameters.

4. The method according to claim 1, wherein the second feature processing network is obtained in the following manner: Based on the second training samples, the initial feature processing network in the second processing model is trained to obtain the second feature processing network; in, The second processing model includes the baseline processing network and the initial feature processing network.

5. The method according to claim 1, wherein training the service processing model constructed by the benchmark processing network and the feature processing network based on the first training samples and the second training samples under the second service type to obtain the service processing model includes: The first training sample is sampled according to a preset ratio to obtain a first sampled sample, and the second training sample is sampled according to the preset ratio to obtain a second sampled sample; Based on the first sample and the second sample, the feature processing network in the service processing network constructed by the benchmark processing network and the feature processing network is trained to obtain the trained service processing model.

6. The method according to claim 1, further comprising: Obtain the pending data of the target service; The data to be processed includes data under the first service type or the second service type; The data to be processed is input into the trained service processing model for service processing to obtain the service processing result.

7. The method according to claim 1, wherein training the initial feature processing network in the first processing model under the first service type based on the first training samples to obtain the first feature processing network comprises: Based on the first training sample, the initial feature output network in the initial feature processing network of the first processing model is trained to obtain the first feature output network; The first feature processing network is constructed based on the feature transformation network and the first feature output network in the initial feature processing network.

8. The method according to claim 7, wherein fusing the first feature processing network and the second feature processing network in the second processing model under the second service type to obtain a feature processing network comprises: The first feature output network and the second feature processing network are merged to obtain a merged output network. The feature processing network is constructed based on the feature transformation network and the merged output network.

9. The method according to claim 8, wherein training the service processing model constructed by the benchmark processing network and the feature processing network based on the first training samples and the second training samples under the second service type to obtain the service processing model includes: Based on the first training samples, the first feature output network in the service processing model to be trained is trained, and based on the second training samples, the second feature output network in the service processing model to be trained is trained to obtain the service processing model.

10. The method of claim 9, further comprising: The data to be processed for the target service is input into the trained service processing model for service processing to obtain the service processing result. The service processing includes: The data to be processed is classified by type to obtain the service type of the data to be processed; The data to be processed is input into a trained feature transformation network to perform feature transformation, thereby obtaining transformed features; The transformation features are input into the feature output network corresponding to the service type for output calculation to obtain the output features; The output features are input into the benchmark processing network for feature processing to obtain the service processing result.

11. A service processing method, comprising: Retrieve the pending data from the target service; The data to be processed includes data to be processed under the first service type or the second service type of the target service. The target service includes a risk identification service. The first service type includes a user risk identification type, and the second service type includes a merchant risk identification type. The data to be processed under the first service type is user data to be identified, and the data to be processed under the second service type is merchant data to be identified. The data to be processed is input into the service processing model for service processing to obtain the service processing result. The service processing model is obtained by training a service processing model constructed from a benchmark processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type. The feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type. The first feature processing network is obtained by training the initial feature processing network in the first processing model under the first service type based on the first training samples. The first processing model includes a benchmark processing network and the initial feature processing network. The second feature processing network is obtained by training the initial feature processing network in the second processing model under the second service type based on the second training samples. The second processing model includes a benchmark processing network and the initial feature processing network.

12. The method of claim 11, wherein the service processing includes: The data to be processed is input into a feature transformation network for feature transformation to obtain transformed features; The transformed features are input into the feature output network for output calculation to obtain the output features; The output features are input into the benchmark processing network for feature processing to obtain the service processing result.

13. The method of claim 11, wherein the service processing includes: The data to be processed is classified by type to obtain the service type of the data to be processed; The data to be processed is input into a feature transformation network for feature transformation to obtain transformed features; The transformation features are input into the feature output network corresponding to the service type for output calculation to obtain the output features; The output features are input into the benchmark processing network for feature processing to obtain the service processing result.

14. A service processing model training apparatus, comprising: The sample determination module is configured to determine a first training sample based on service data under a first service type of the target service, wherein the target service includes a risk identification service, the first service type includes a user risk identification type, and the first training sample includes user data and risk categories corresponding to each user data. The network training module is configured to train the initial feature processing network in the first processing model under the first service type based on the first training samples, so as to obtain the first feature processing network. The first processing model includes a baseline processing network and the initial feature processing network; The fusion processing module is configured to fuse the first feature processing network and the second feature processing network in the second processing model under the second service type of the target service to obtain a feature processing network. The second feature processing network is obtained by training the initial feature processing network in the second processing model based on the second training samples under the second service type. The first service type and the second service type represent different service types under the target service. The second service type includes merchant risk identification type. The second training samples include merchant data and the risk category corresponding to each merchant data. The model training module is configured to train the service processing model constructed by the benchmark processing network and the feature processing network based on the first training samples and the second training samples under the second service type, so as to obtain the service processing model.

15. A service processing apparatus, comprising: The data acquisition module is configured to acquire the data to be processed from the target service; The data to be processed includes data to be processed under the first service type or the second service type of the target service. The target service includes a risk identification service. The first service type includes a user risk identification type, and the second service type includes a merchant risk identification type. The data to be processed under the first service type is user data to be identified, and the data to be processed under the second service type is merchant data to be identified. The service processing module is configured to input the data to be processed into the service processing model for service processing and obtain the service processing result. The service processing model is obtained by training a service processing model constructed from a benchmark processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type. The feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type. The first feature processing network is obtained by training the initial feature processing network in the first processing model under the first service type based on the first training samples. The first processing model includes a benchmark processing network and the initial feature processing network. The second feature processing network is obtained by training the initial feature processing network in the second processing model under the second service type based on the second training samples. The second processing model includes a benchmark processing network and the initial feature processing network.

16. A service processing model training device, comprising: processor; And, a memory configured to store computer-executable instructions, which, when executed, cause the processor to: The first training sample is determined based on the service data under the first service type of the target service, wherein the target service includes risk identification service, the first service type includes user risk identification type, and the first training sample includes user data and risk categories corresponding to each user data. Based on the first training sample, the initial feature processing network in the first processing model under the first service type is trained to obtain the first feature processing network; The first processing model includes a baseline processing network and the initial feature processing network; The first feature processing network and the second feature processing network in the second processing model under the second service type of the target service are fused to obtain a feature processing network. The second feature processing network is obtained by training the initial feature processing network in the second processing model based on the second training samples under the second service type. The first service type and the second service type represent different service types under the target service. The second service type includes merchant risk identification type. The second training samples include merchant data and the risk category corresponding to each merchant data. Based on the first training samples and the second training samples under the second service type, the service processing model constructed by the benchmark processing network and the feature processing network is trained to obtain the service processing model.

17. A service processing device, comprising: processor; And, a memory configured to store computer-executable instructions, which, when executed, cause the processor to: Retrieve the pending data from the target service; The data to be processed includes data to be processed under the first service type or the second service type of the target service. The target service includes a risk identification service. The first service type includes a user risk identification type, and the second service type includes a merchant risk identification type. The data to be processed under the first service type is user data to be identified, and the data to be processed under the second service type is merchant data to be identified. The data to be processed is input into the service processing model for service processing to obtain the service processing result. The service processing model is obtained by training a service processing model constructed from a benchmark processing network and a feature processing network based on a first training sample under the first service type and a second training sample under the second service type. The feature processing network is obtained by integrating the first feature processing network under the first service type and the second feature processing network under the second service type. The first feature processing network is obtained by training the initial feature processing network in the first processing model under the first service type based on the first training samples. The first processing model includes a benchmark processing network and the initial feature processing network. The second feature processing network is obtained by training the initial feature processing network in the second processing model under the second service type based on the second training samples. The second processing model includes a benchmark processing network and the initial feature processing network.

Images