CN112418443A - Data processing method, device and equipment based on transfer learning and storage medium - Google Patents

Data processing method, device and equipment based on transfer learning and storage medium Download PDF

Info

Publication number
CN112418443A
CN112418443A CN202011393272.4A CN202011393272A CN112418443A CN 112418443 A CN112418443 A CN 112418443A CN 202011393272 A CN202011393272 A CN 202011393272A CN 112418443 A CN112418443 A CN 112418443A
Authority
CN
China
Prior art keywords
model
sample
feature
loss value
prediction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011393272.4A
Other languages
Chinese (zh)
Inventor
康焱
刘洋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
WeBank Co Ltd
Original Assignee
WeBank Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by WeBank Co Ltd filed Critical WeBank Co Ltd
Priority to CN202011393272.4A priority Critical patent/CN112418443A/en
Publication of CN112418443A publication Critical patent/CN112418443A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a data processing method, a device, equipment and a storage medium based on transfer learning, wherein the method comprises the following steps: obtaining a first sample feature set and obtaining a second sample feature set; determining a first sample feature characterization based on the first sample feature set and a second sample feature characterization based on the second sample feature set; updating the feature extraction model based on each first sample feature characterization and each second sample feature characterization; and determining a target prediction model and a target feature extraction model based on the first sample feature characterization and the interpretable prediction model to be trained. According to the method, a target model which gives consideration to model migration and model interpretability can be obtained, the knowledge of the sample is migrated through the target feature extraction model, the efficiency of migration learning is improved through the domain distinguishing model in the migration learning process, and the efficiency of data processing and the utilization rate of computer computing resources are improved.

Description

Data processing method, device and equipment based on transfer learning and storage medium
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a data processing method, a data processing device, data processing equipment and a storage medium based on transfer learning.
Background
Machine learning performs supervised training on a large amount of labeled data to achieve good performance and effect, however, a large labeled data set is limited in quantity and application field, and manually labeling a sufficient amount of training data is often costly. Aiming at the problem, a transfer learning method is generally adopted to solve the problem, namely, a discriminator is trained to adjust parameters of a transfer learning network, so that the distribution deviation between data in a source field and data in a target field is reduced under the transfer learning network after the parameters are adjusted, and the transfer learning network has a better effect when the target field is applied to finish a target task.
However, the lack of interpretability of deep learning models makes them difficult to use for migratory learning in applications that require model interpretability (e.g., financial risk control), and low complexity deep learning models have a weak ability to learn migratable knowledge from raw data and thus are not robust in migratory ability. This creates a contradiction, the deep learning model with strong migration capability lacks interpretability, and the deep learning model with strong interpretability has weak migration capability, so that the deep learning model cannot take account of both interpretability and migration capability. Meanwhile, in the transfer learning, because the data of different fields or scenes cannot be effectively distinguished, the trained model can reach the target performance only after being trained for a longer time in the transfer learning process, so that the transfer learning efficiency is low, a computer needs to consume a large amount of resources and computing power, and the utilization rate of computer computing power resources is low.
The above is only for the purpose of assisting understanding of the technical aspects of the present invention, and does not represent an admission that the above is prior art.
Disclosure of Invention
The invention mainly aims to provide a data processing method, a data processing device, data processing equipment and a data processing storage medium based on transfer learning, and aims to solve the technical problems that the deep learning model cannot give consideration to interpretability and transfer capability, and the transfer learning efficiency is low, so that the utilization rate of computer power resources is low.
In order to achieve the above object, the present invention provides a data processing method based on transfer learning, which includes the following steps:
grouping the characteristics of first samples of a first service scene based on preset service requirements to obtain a preset number of first sample characteristic groups, and grouping the characteristics of second samples of a second service scene based on the preset service requirements to obtain a preset number of second sample characteristic groups;
determining a first sample feature characterization based on each first sample feature group and the corresponding feature extraction model thereof, and determining a second sample feature characterization based on each second sample feature group and the corresponding feature extraction model thereof;
determining a first domain distinguishing loss value based on each first sample feature characterization and a corresponding domain distinguishing model thereof, determining a second domain distinguishing loss value based on each second sample feature characterization and a corresponding domain distinguishing model thereof, and updating each feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value;
and determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
Further, the step of determining a first domain discrimination loss value based on each first sample feature representation and its corresponding domain discrimination model comprises:
inputting each first sample characteristic representation into a corresponding domain distinguishing model to obtain a first domain distinguishing loss value corresponding to each first sample characteristic representation;
and updating the corresponding domain distinguishing model based on the first domain distinguishing loss value to obtain an updated domain distinguishing model, and updating the corresponding feature extraction model of each first sample feature characterization based on each first domain distinguishing loss value through domain confrontation learning to obtain an intermediate feature extraction model.
Further, the step of determining a second domain discrimination loss value based on each second sample feature characterization and the corresponding domain discrimination model thereof, and updating each feature extraction model based on the first domain discrimination loss value and the second domain discrimination loss value includes:
inputting each second sample characteristic representation into the corresponding updated domain distinguishing model to obtain a second domain distinguishing loss value corresponding to each second sample characteristic representation;
and updating the intermediate feature extraction model corresponding to the feature characterization of each second sample through domain confrontation learning based on the distinguishing loss value of each second domain to obtain the updated feature extraction model.
Further, the step of determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model comprises:
inputting the characteristic feature set corresponding to the first sample characteristic representation into a prediction model to be trained for model training to obtain a prediction loss value;
updating the prediction model to be trained based on the prediction loss value to obtain a target prediction model;
and updating each updated feature extraction model based on the predicted loss value to obtain a target feature extraction model.
Further, the step of grouping the features of the first sample of the first service scenario based on the preset service requirement to obtain a preset number of first sample feature groups includes:
grouping the characteristics of the first sample based on preset service requirements to obtain a first sample characteristic group corresponding to the first sample and a plurality of single characteristics;
the step of inputting the feature characteristic set corresponding to the first sample feature characterization into the prediction model to be trained for model training to obtain the prediction loss value comprises:
and inputting the first sample characteristic representation and the characteristic feature set corresponding to the monomer characteristic into a prediction model to be trained for model training to obtain the prediction loss value.
Further, the target prediction model is a fraud score prediction model; the first service scene comprises a credit scoring scene of a user, the second service scene comprises a fraud scoring scene of the user, the first sample comprises credit scoring data in a first participant, and the second sample is fraud scoring data in the first participant;
the step of determining a first domain discrimination loss value based on each first sample feature characterization and the corresponding domain discrimination model, determining a second domain discrimination loss value based on each second sample feature characterization and the corresponding domain discrimination model, and updating each feature extraction model based on the first domain discrimination loss value and the second domain discrimination loss value comprises:
determining a first domain distinguishing loss value based on a first sample feature representation corresponding to credit scoring data in a first participant and a domain distinguishing model corresponding to the first sample feature representation, determining a second domain distinguishing loss value based on a second sample feature representation corresponding to fraud scoring data in the first participant and a domain distinguishing model corresponding to the fraud scoring data in the first participant, and updating each feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value;
the step of determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model comprises the following steps:
and determining a prediction loss value and a fraud score prediction model based on a first sample feature representation corresponding to credit score data in a first participant and an interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
Further, after the steps of determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model, the data processing method based on the transfer learning further includes:
grouping samples to be predicted of a second service scene based on preset service requirements to obtain a sample feature group to be predicted;
inputting a sample feature group to be predicted into a corresponding target feature extraction model to obtain a plurality of feature representations to be predicted;
and inputting the feature characterization set corresponding to the feature characterization to be predicted into the target prediction model to obtain a prediction result.
Further, the step of grouping the to-be-predicted samples of the second service scenario based on the preset service requirement to obtain the to-be-predicted sample feature set includes:
grouping samples to be predicted of a second service scene based on preset service requirements to obtain a sample feature group to be predicted and a plurality of monomer features to be predicted;
the step of inputting the feature characterization set corresponding to the feature characterization to be predicted into the target prediction model to obtain the prediction result comprises the following steps:
and inputting the feature characterization to be predicted and the feature characterization set corresponding to the monomer feature to be predicted into the target prediction model to obtain a prediction result.
Further, to achieve the above object, the present invention provides a data processing apparatus based on migration learning, comprising:
the grouping module is used for grouping the characteristics of the first samples of the first service scene based on preset service requirements to obtain a preset number of first sample characteristic groups, and grouping the characteristics of the second samples of the second service scene based on the preset service requirements to obtain a preset number of second sample characteristic groups;
the characteristic extraction module is used for determining first sample characteristic characteristics based on each first sample characteristic group and the corresponding characteristic extraction model thereof, and determining second sample characteristic characteristics based on each second sample characteristic group and the corresponding characteristic extraction model thereof;
the model determining module is used for determining a first domain distinguishing loss value based on each first sample characteristic characterization and the corresponding domain distinguishing model thereof, determining a second domain distinguishing loss value based on each second sample characteristic characterization and the corresponding domain distinguishing model thereof, and updating each feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value;
and the prediction model determining module is used for determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
Further, to achieve the above object, the present invention also provides a data processing apparatus based on migration learning, comprising: the data processing method comprises a memory, a processor and a data processing program based on the transfer learning, wherein the data processing program based on the transfer learning is stored on the memory and can run on the processor, and when being executed by the processor, the data processing program based on the transfer learning realizes the steps of the data processing method based on the transfer learning.
In addition, to achieve the above object, the present invention further provides a storage medium having a data processing program based on the transfer learning stored thereon, wherein the data processing program based on the transfer learning realizes the steps of the data processing method based on the transfer learning when being executed by a processor.
In addition, to achieve the above object, the present invention further provides a computer program product, which when executed, realizes the steps of the foregoing data processing method based on the transfer learning.
The method comprises the steps of grouping the characteristics of first samples of a first service scene based on preset service requirements to obtain a preset number of first sample characteristic groups, and grouping the characteristics of second samples of a second service scene based on the preset service requirements to obtain a preset number of second sample characteristic groups; determining a first sample feature characterization based on each first sample feature group and the corresponding feature extraction model thereof, and determining a second sample feature characterization based on each second sample feature group and the corresponding feature extraction model thereof; then, determining a first domain distinguishing loss value based on each first sample characteristic characterization and a corresponding domain distinguishing model thereof, determining a second domain distinguishing loss value based on each second sample characteristic characterization and a corresponding domain distinguishing model thereof, and updating each characteristic extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value; and then determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, determining a target feature extraction model based on the prediction loss value and the updated feature extraction model, obtaining a target model which gives consideration to both model migration and model interpretable, migrating the knowledge of the sample through the target feature extraction model, predicting through the interpretable target prediction model, and explaining the contribution of each feature group to the prediction result, thereby achieving the purpose of giving consideration to both model migration and model interpretable. Meanwhile, the first sample and the second sample are accurately distinguished by the field distinguishing model in the transfer learning process, so that the transfer learning efficiency is improved, the defect that a large amount of computer computing resources are consumed in the prior art is overcome, and the data processing efficiency and the utilization rate of the computer computing resources are improved.
Drawings
FIG. 1 is a schematic diagram of a data processing apparatus based on transfer learning in a hardware operating environment according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating a data processing method based on transfer learning according to a first embodiment of the present invention;
FIG. 3 is a schematic diagram of a first sample and a second sample in the data processing method based on the transfer learning according to the present invention;
FIG. 4 is a schematic diagram of a training process of transfer learning in the data processing method based on transfer learning according to the present invention;
FIG. 5 is a schematic diagram of a prediction process in the data processing method based on transfer learning according to the present invention;
FIG. 6 is a functional block diagram of an embodiment of a data processing apparatus based on transfer learning according to the present invention.
The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.
Detailed Description
It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1, fig. 1 is a schematic structural diagram of a data processing device based on migration learning in a hardware operating environment according to an embodiment of the present invention.
The data processing device based on the migration learning in the embodiment of the present invention may be a PC, or may be a mobile terminal device having a display function, such as a smart phone, a tablet computer, an electronic book reader, an MP3(Moving Picture Experts Group Audio Layer III, motion video Experts compression standard Audio Layer 3) player, an MP4(Moving Picture Experts Group Audio Layer IV, motion video Experts compression standard Audio Layer 4) player, a portable computer, and the like.
As shown in fig. 1, the data processing apparatus based on the transfer learning may include: a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, a communication bus 1002. Wherein a communication bus 1002 is used to enable connective communication between these components. The user interface 1003 may include a Display screen (Display), an input unit such as a Keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface). The memory 1005 may be a high-speed RAM memory or a non-volatile memory (e.g., a magnetic disk memory). The memory 1005 may alternatively be a storage device separate from the processor 1001.
Optionally, the data processing device based on the transfer learning may further include a camera, a Radio Frequency (RF) circuit, a sensor, an audio circuit, a WiFi module, and the like. Such as light sensors, motion sensors, and other sensors. Of course, the data processing device based on the migration learning may also be configured with other sensors such as a gyroscope, a barometer, a hygrometer, a thermometer, and an infrared sensor, which are not described herein again.
Those skilled in the art will appreciate that the terminal architecture shown in fig. 1 does not constitute a limitation of a data processing apparatus based on migration learning, and may include more or fewer components than those shown, or some components in combination, or a different arrangement of components.
As shown in fig. 1, a memory 1005, which is a kind of computer storage medium, may include therein an operating system, a network communication module, a user interface module, and a data processing program based on the migration learning.
In the data processing device based on the transfer learning shown in fig. 1, the network interface 1004 is mainly used for connecting with a background server and performing data communication with the background server; the user interface 1003 is mainly used for connecting a client (user side) and performing data communication with the client; and the processor 1001 may be used to invoke a migration learning based data processing program stored in the memory 1005.
In the present embodiment, the data processing apparatus based on the migration learning includes: the data processing method comprises a memory 1005, a processor 1001 and a data processing program based on the transfer learning, wherein the data processing program based on the transfer learning is stored in the memory 1005 and can be run on the processor 1001, and when the processor 1001 calls the data processing program based on the transfer learning stored in the memory 1005, the steps of the data processing method based on the transfer learning in the following embodiments are executed.
The invention also provides a data processing method based on the transfer learning, and referring to fig. 2, fig. 2 is a schematic flow diagram of a first embodiment of the data processing method based on the transfer learning.
Step S101, grouping the characteristics of first samples of a first service scene based on preset service requirements to obtain a preset number of first sample characteristic groups, and grouping the characteristics of second samples of a second service scene based on the preset service requirements to obtain a preset number of second sample characteristic groups;
in this embodiment, the first service scenario and the second service scenario are two service scenarios with similar but different services, for example, the first service scenario is credit score and fraud score, and the second service scenario is fraud score; the first service scenario has a large number of samples, and the second service scenario has a certain number of samples, that is, the number of the first samples is (far) greater than the number of the second samples; each sample in the first sample corresponds to a category label and a domain label, each sample in the second sample corresponds to a domain label, the domain label of the first sample is different from the domain label of the second sample, for example, the domain label corresponds to a business scenario, the domain label of each sample in the first sample is 1, and the domain label of each sample in the second sample is 0.
In this embodiment, a first sample of a first service scenario and a second sample of a second service scenario are obtained, in order to make the number of first feature characterizations the same as the number of second feature characterizations, for the first service scenario and the second service scenario, a part of features that are completely the same between the first service scenario and the second service scenario are determined as preset service requirements, then the features of the first sample and the second sample are grouped based on the preset service requirements, specifically, the features of the first sample of the first service scenario are grouped based on the preset service requirements to obtain a preset number of first sample feature groups, and the features of the second sample of the second service scenario are grouped based on the preset service requirements to obtain a preset number of second sample feature groups, so that the features in each group of samples are descriptions of a certain attribute of the samples, and each sample group can look at a comparative abstract description of a certain aspect of the samples, for example, dating, microblog dating, WeChat dating, etc. may be grouped together as a set of characteristics that are descriptive of the social attributes of the sample. The preset number is reasonably set according to preset service requirements.
Referring to fig. 3, in fig. 3, a service a is a first service scenario, a service B is a second service scenario, and Fi AIs a first sample feature group, Fi BIs a second sample feature set, fj AIs a monomer characteristic corresponding to the first sample, fj BThe corresponding monomer characteristic of the second sample.
Step S102, determining first sample feature characteristics based on each first sample feature group and the corresponding feature extraction model thereof, and determining second sample feature characteristics based on each second sample feature group and the corresponding feature extraction model thereof;
in this embodiment, after a first sample feature group and a second sample feature group are obtained, a first sample feature characterization is determined based on each first sample feature group and a corresponding feature extraction model thereof, and a second sample feature characterization is determined based on each second sample feature group and a corresponding feature extraction model thereof, specifically, each first sample feature group is input into a corresponding feature extraction model thereof for feature extraction to obtain a first sample feature characterization corresponding to each first sample feature group, each second sample feature group is input into a corresponding feature extraction model thereof for feature extraction to obtain a second sample feature characterization corresponding to each second sample feature group, for example, the first sample feature group includes n groups of features such as feature group 1, feature group 2 … …, and the second sample feature group also includes feature group 1, feature group n, and the like, The feature group 2 … … includes n sets of features such as the feature group n, and the feature extraction model includes the feature extraction model 1 and the feature extraction model 2 … …, and the same feature group is required to correspond to the same required feature extraction model, that is, the feature group 1 of the first sample feature group and the feature group 1 of the second sample feature group both correspond to the feature extraction model 1, and the feature group n of the first sample feature group and the feature group n of the second sample feature group both correspond to the feature extraction model n.
Step S103, determining a first domain distinguishing loss value based on each first sample characteristic characterization and a corresponding domain distinguishing model thereof, determining a second domain distinguishing loss value based on each second sample characteristic characterization and a corresponding domain distinguishing model thereof, and updating each feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value;
in this embodiment, when the first feature representations and the second feature representations are obtained, a plurality of first domain discrimination loss values are determined based on each first feature representation and the corresponding domain discrimination model thereof, specifically, the first feature representations are input into the corresponding domain discrimination models thereof to obtain first prediction domain labels corresponding to the first feature representations, and the first domain discrimination loss values corresponding to the first feature representations are calculated according to the first prediction domain labels and the domain labels corresponding to the first feature representations, respectively. And determining a plurality of second domain distinguishing loss values based on each second feature characterization and the corresponding domain distinguishing model thereof, specifically, inputting each second feature characterization into the corresponding domain distinguishing model thereof to obtain a second prediction domain label corresponding to each second feature characterization, and calculating the second domain distinguishing loss value corresponding to each second feature characterization according to the second prediction domain label and the domain label corresponding to each second feature characterization. And then, updating each corresponding feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value to obtain an updated feature extraction model.
And step S104, determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
In this embodiment, after the updated feature extraction model is determined, a prediction loss value and a target prediction model are determined based on a first feature representation and an interpretable prediction model to be trained, specifically, a feature representation set corresponding to the first feature representation is input into the interpretable prediction model to be trained for model training to obtain a prediction loss value, the feature representation set corresponding to the first feature representation includes all first feature representation data sets, the prediction model to be trained is updated according to the prediction loss value to obtain the target prediction model, and the updated feature extraction model is updated according to the first prediction loss value to obtain the target feature extraction model.
It should be noted that, in other embodiments, when determining the target feature extraction model, the number of training rounds of the prediction model to be trained is accumulated, and if the number of training rounds is smaller than a preset value, the target prediction model is used as the prediction model to be trained, the target feature extraction model is used as the updated feature extraction model, and the step S104 is continuously executed.
The prediction model to be trained is an interpretable machine learning model, for example, the prediction model to be trained is a linear regression model or a logistic regression model.
However, the design is not limited thereto, and in other embodiments, the target prediction model is a fraud score prediction model; the first service scene comprises a credit scoring scene of a user, the second service scene comprises a fraud scoring scene of the user, the first sample comprises credit scoring data in a first participant, and the second sample is fraud scoring data in the first participant;
the step S103 comprises the steps of determining a first domain distinguishing loss value based on a first sample feature representation corresponding to credit scoring data in a first participant and a corresponding domain distinguishing model thereof, determining a second domain distinguishing loss value based on a second sample feature representation corresponding to fraud scoring data in the first participant and a corresponding domain distinguishing model thereof, and updating each feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value;
step S104 includes: and determining a prediction loss value and a fraud score prediction model based on a first sample feature representation corresponding to credit score data in a first participant and an interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
Specifically, a first sample feature representation corresponding to credit scoring data in a first participant is input into a corresponding field distinguishing model to obtain a first field distinguishing loss value, a second sample feature representation corresponding to fraud scoring data in the first participant is input into the corresponding field distinguishing model to obtain a second field distinguishing loss value, and then each feature extraction model is updated based on the first field distinguishing loss value and the second field distinguishing loss value to obtain an updated feature extraction model.
Then, inputting the first sample feature representation corresponding to the credit scoring data in the first participant into an interpretable prediction model to be trained for model training to obtain a prediction loss value, updating the interpretable prediction model to be trained according to the prediction loss value to obtain a fraud scoring prediction model, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
In the data processing method based on the transfer learning provided in this embodiment, the features of the first samples of the first service scenario are grouped based on the preset service requirement to obtain a preset number of first sample feature groups, and the features of the second samples of the second service scenario are grouped based on the preset service requirement to obtain a preset number of second sample feature groups; determining a first sample feature characterization based on each first sample feature group and the corresponding feature extraction model thereof, and determining a second sample feature characterization based on each second sample feature group and the corresponding feature extraction model thereof; then, determining a first domain distinguishing loss value based on each first sample characteristic characterization and a corresponding domain distinguishing model thereof, determining a second domain distinguishing loss value based on each second sample characteristic characterization and a corresponding domain distinguishing model thereof, and updating each characteristic extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value; and then determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, determining a target feature extraction model based on the prediction loss value and the updated feature extraction model, obtaining a target model which gives consideration to both model migration and model interpretable, migrating the knowledge of the sample through the target feature extraction model, predicting through the interpretable target prediction model, and explaining the contribution of each feature group to the prediction result, thereby achieving the purpose of giving consideration to both model migration and model interpretable. Meanwhile, the first sample and the second sample are accurately distinguished by the field distinguishing model in the transfer learning process, so that the transfer learning efficiency is improved, the defect that a large amount of computer computing resources are consumed in the prior art is overcome, and the data processing efficiency and the utilization rate of the computer computing resources are improved.
A second embodiment of the data processing method based on the migration learning of the present invention is proposed based on the first embodiment, and in this embodiment, the step S103 includes:
step S201, inputting each first sample feature into a corresponding domain distinguishing model to obtain a first domain distinguishing loss value corresponding to each first sample feature;
step S202, updating the corresponding domain distinguishing model based on the first domain distinguishing loss value to obtain an updated domain distinguishing model, and updating the feature extraction model corresponding to each first sample feature characterization based on each first domain distinguishing loss value through domain confrontation learning to obtain an intermediate feature extraction model.
In this embodiment, when the first feature representations are obtained, the domain distinguishing models corresponding to the input of the first feature representations are subjected to model training to obtain first domain distinguishing loss values corresponding to the first feature representations, that is, the first feature representations are input into the corresponding domain distinguishing models respectively to obtain first prediction domain labels corresponding to the first feature representations, and the first domain distinguishing loss values corresponding to the first feature representations are calculated according to the first prediction domain labels and the domain labels corresponding to the first feature representations respectively.
And then, updating the domain distinguishing model corresponding to each first characteristic representation according to each first domain distinguishing loss value to obtain an updated domain distinguishing model, and updating the characteristic extraction model corresponding to each first characteristic representation through domain confrontation learning based on each first domain distinguishing loss value to obtain an intermediate characteristic extraction model.
For example, the first characterization includes ra1、ra2……ranThe domain-specific model includes n, D1-DnThen r will bea1Input D1，ra2Input D, ranInputting Dn, and respectively carrying out model training to obtain n first prediction field labels By1、By2……BynAccording to ra1、ra2……ranCorresponding (real) domain label Bs1、Bs2……BsnAnd By1、By2……BynRespectively determining corresponding first domain distinguishing loss values L1、L2……LnI.e. according to By1、Bs1Determination of L1According to By2、Bs2Determination of L2According to Byn、BsnDetermining Ln and then based on L1Update D1According to L2Update D2According to L1Update D1According to LnUpdate DnObtaining each updated domain distinguishing model, and according to L respectively1、L2……LnUpdate each ra1、ra2……ranCorresponding feature extraction models, i.e. according to L1Update ra1Corresponding feature extraction model according to L2Update ra2Corresponding feature extraction model according to L1Update D1According to LnUpdate ranAnd obtaining each intermediate characteristic extraction model by the corresponding characteristic extraction model.
In the data processing method based on the transfer learning provided in this embodiment, each first sample feature representation is input into a corresponding domain distinguishing model, so as to obtain a first domain distinguishing loss value corresponding to each first sample feature representation; and then updating the corresponding domain distinguishing model based on the first domain distinguishing loss value to obtain an updated domain distinguishing model, updating the feature extraction model corresponding to each first sample feature characterization based on each first domain distinguishing loss value through domain confrontation learning to obtain an intermediate feature extraction model, realizing updating of the domain distinguishing model and the feature extraction model through the first sample feature characterization, and improving the accuracy of model training.
A third embodiment of the data processing method based on the migration learning of the present invention is proposed based on the second embodiment, and in this embodiment, step S103 includes:
step S301, inputting each second feature representation into the corresponding updated domain distinguishing model to obtain a second domain distinguishing loss value corresponding to each second feature representation;
step S302, updating the intermediate feature extraction model corresponding to each second sample feature characterization through domain confrontation learning based on each second domain distinguishing loss value so as to obtain an updated feature extraction model.
In this embodiment, each second feature representation is input into the corresponding updated domain distinguishing model for model training, so as to obtain a second domain distinguishing loss value corresponding to each second feature representation, that is, the second feature representations are respectively input into the corresponding updated domain distinguishing models, so as to obtain a second prediction domain label corresponding to each second feature representation, and the second domain distinguishing loss value corresponding to each second feature representation is calculated according to the second prediction domain label and the domain label corresponding to each second feature representation.
And then, updating the intermediate feature extraction model corresponding to each second feature characterization through domain confrontation learning based on each second domain distinguishing loss value so as to obtain an updated feature extraction model. In other embodiments, the updated domain distinguishing model corresponding to each second feature representation may be updated according to each second domain distinguishing loss value to obtain a target domain distinguishing model, so that the target domain distinguishing model is used as the domain distinguishing model during subsequent training.
In the data processing method based on the transfer learning provided in this embodiment, each second sample feature representation is input into the corresponding updated domain distinguishing model, so as to obtain a second domain distinguishing loss value corresponding to each second sample feature representation; and then updating the intermediate feature extraction model corresponding to each second sample feature characterization through field confrontation learning based on each second field distinguishing loss value to obtain an updated feature extraction model, updating the feature extraction model through the second sample feature characterization, and improving the accuracy of model training.
A fourth embodiment of the data processing method based on the migration learning of the present invention is proposed based on the first embodiment, and in this embodiment, the step S104 includes:
step S401, inputting a feature characterization set corresponding to the first feature characterization into the prediction model to be trained for model training to obtain a prediction loss value;
step S402, updating the prediction model to be trained based on the prediction loss value to obtain a target prediction model;
step S403, updating each updated feature extraction model based on the predicted loss value to obtain a target feature extraction model.
In this embodiment, after the updated feature extraction model is determined, the feature set corresponding to the first feature representation is input to the prediction model to be trained for model training to obtain a prediction loss value, that is, the feature set corresponding to the first feature representation is obtained according to the serial number sequence of the first feature representation, the feature set corresponding to the first feature representation is a data set including all the first feature representations, the feature set corresponding to the first feature representation is input to the prediction model to be trained for model training to obtain a prediction category label corresponding to the first feature representation, and the prediction loss value is calculated according to the prediction category label corresponding to the first feature representation and the category label (real category label) corresponding to the first feature representation.
And then, updating the prediction model to be trained according to the prediction loss value to obtain a target prediction model, and updating the updated feature extraction model according to the prediction loss value to obtain a target feature extraction model.
Further, in one embodiment,
step S101 includes: grouping the characteristics of the first sample based on preset service requirements to obtain a first sample characteristic group corresponding to the first sample and a plurality of single characteristics;
step S401 includes: and inputting the first sample characteristic representation and the characteristic feature set corresponding to the monomer characteristic into a prediction model to be trained for model training to obtain the prediction loss value.
In this embodiment, when the features of the first sample are grouped, there are often samples that are not grouped, and such samples are monomer features, and when the prediction model is trained, in order to improve the accuracy of the model, the first feature characterization and the feature set corresponding to the monomer features are input into the prediction model to be trained for model training, so as to obtain a prediction loss value, that is, the first feature characterization and the monomer features are first spliced into one data set in sequence, the data set is input into the prediction model to be trained for model training, so as to obtain a corresponding prediction class label, and the prediction loss value is calculated according to the prediction class label and the corresponding class label (true class label), so as to improve the accuracy of the model training.
Referring to FIG. 4, in FIG. 4, the feature group F in the first sample feature group is passed1 AAnd a feature group F in the second sample feature group1 BTraining to obtain a target characteristic model R1… … passing through feature set F in the first sample feature setn AAnd a feature group F in the second sample feature groupn BTraining to obtain an updated feature model Rn，D1……D3For a domain differentiation model, Ld,1……Ld,3The loss values are distinguished for the domain. Grouping the first sample features Fi AAnd a plurality of monomer characteristics fj AInputting a prediction model to be trained to obtain a prediction loss value Lcls AFinally based on the predicted loss value Lcls AAnd updating each updated feature extraction model to obtain a target feature extraction model, and updating the to-be-trained prediction model to obtain a target prediction model.
In the data processing method based on the transfer learning provided by this embodiment, the feature set corresponding to the first sample feature representation is input to a prediction model to be trained for model training, so as to obtain a prediction loss value; then updating the prediction model to be trained based on the prediction loss value to obtain a target prediction model; and updating each updated feature extraction model based on the prediction loss value to obtain a target feature extraction model, and updating the prediction model to be trained and the updated feature extraction model through the first feature characterization to improve the accuracy of the target prediction model and the target feature extraction model.
A fifth embodiment of the data processing method based on the migration learning according to the present invention is proposed based on the above embodiments, and in this embodiment, after step S104, the method further includes:
step S501, grouping the characteristics of the samples to be predicted of the second service scene based on the preset service requirements to obtain a sample characteristic group to be predicted;
step S502, inputting a sample feature group to be predicted into a corresponding target feature extraction model to obtain a plurality of feature representations to be predicted;
and S503, inputting a feature characterization set corresponding to the feature characterization to be predicted into the target prediction model to obtain a prediction result.
In this embodiment, when prediction is required, a to-be-predicted sample of a second service scene is obtained, then features of the to-be-predicted sample are grouped based on preset service requirements to obtain a to-be-predicted sample feature group, the to-be-predicted sample feature group is input into a corresponding target feature extraction model to obtain a plurality of to-be-predicted feature characterizations, wherein the number/sequence of the plurality of to-be-predicted sample feature groups is the same as that of the first sample group, each to-be-predicted sample feature group is input into the corresponding target feature extraction model to obtain a plurality of to-be-predicted feature characterizations, that is, each to-be-predicted sample feature group corresponds to one to-be-predicted feature characterization.
And then, splicing the feature characterizations to be predicted to obtain a feature characterization set corresponding to the feature characterizations to be predicted, and inputting the feature characterization set corresponding to the feature characterizations to be predicted into a target prediction model to obtain a prediction result, namely the pre-stored type of the sample to be predicted.
Further, in one embodiment,
step 501 comprises: grouping the characteristics of the samples to be predicted of the second service scene based on preset service requirements to obtain a sample characteristic group to be predicted and a plurality of monomer characteristics to be predicted;
step 503 comprises: and inputting the feature characterization to be predicted and the feature characterization set corresponding to the monomer feature to be predicted into the target prediction model to obtain a prediction result.
In this embodiment, when the samples to be predicted are grouped, if a sample feature group to be predicted and a plurality of monomer features to be predicted are obtained, the feature features to be predicted and the monomer features to be predicted are spliced to obtain a feature set to be predicted, and the feature set to be predicted is input into the target prediction model to obtain a prediction result, so that the pre-stored category of the sample to be predicted is obtained.
Referring to FIG. 5, in FIG. 5, F1 B、F2 B、F3 BFor the sample to be predicted corresponding to the sample to be predicted, f1 B、f2 B、f3 BFor the monomer characteristics to be predicted of the sample to be predicted, F1 B、F2 B、F3 BRespectively inputting corresponding target feature extraction models R1、R2、R3And then inputting the plurality of characteristics to be predicted and the characteristics of the single body to be predicted into the target prediction model G to obtain a prediction result (namely a class label).
In the data processing method based on the transfer learning provided by the embodiment, the features of the samples to be predicted of the second service scene are grouped based on the preset service requirements, so as to obtain a sample feature group to be predicted; inputting the sample feature group to be predicted into a corresponding target feature extraction model to obtain a plurality of feature representations to be predicted; and then inputting the feature characterization set corresponding to the feature characterization to be predicted into the target prediction model to obtain a prediction result, so that the prediction of the sample to be predicted through the target model is realized, the accuracy of the prediction result can be improved through the target feature extraction model obtained through migration learning, and the reason for obtaining the prediction result can be explained through the target prediction model so as to realize the consideration of model migration and model interpretability.
An embodiment of the present invention further provides a data processing apparatus based on transfer learning, referring to fig. 6, where fig. 6 is a schematic diagram of functional modules of an embodiment of the data processing apparatus based on transfer learning according to the present invention, and the data processing apparatus based on transfer learning includes:
the grouping module 100 is configured to group features of first samples of a first service scenario based on a preset service requirement to obtain a preset number of first sample feature groups, and group features of second samples of a second service scenario based on the preset service requirement to obtain a preset number of second sample feature groups;
a representation extraction module 200, configured to determine a first sample feature representation based on each first sample feature group and the feature extraction model corresponding to the first sample feature group, and determine a second sample feature representation based on each second sample feature group and the feature extraction model corresponding to the second sample feature group;
a model determining module 300, configured to determine a first domain discrimination loss value based on each first sample feature characterization and a corresponding domain discrimination model thereof, determine a second domain discrimination loss value based on each second sample feature characterization and a corresponding domain discrimination model thereof, and update each feature extraction model based on the first domain discrimination loss value and the second domain discrimination loss value;
a prediction model determining module 400, configured to determine a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determine a target feature extraction model based on the prediction loss value and the updated feature extraction model.
Optionally, the model determining module 300 is further configured to:
inputting each first sample characteristic representation into a corresponding domain distinguishing model to obtain a first domain distinguishing loss value corresponding to each first sample characteristic representation;
and updating the corresponding domain distinguishing model based on the first domain distinguishing loss value to obtain an updated domain distinguishing model, and updating the corresponding feature extraction model of each first sample feature characterization based on each first domain distinguishing loss value through domain confrontation learning to obtain an intermediate feature extraction model.
Optionally, the model determining module 300 is further configured to:
inputting each second sample characteristic representation into the corresponding updated domain distinguishing model to obtain a second domain distinguishing loss value corresponding to each second sample characteristic representation;
and updating the intermediate feature extraction model corresponding to the feature characterization of each second sample through domain confrontation learning based on the distinguishing loss value of each second domain to obtain the updated feature extraction model.
Optionally, the prediction model determining module 400 is further configured to:
inputting the characteristic feature set corresponding to the first sample characteristic representation into a prediction model to be trained for model training to obtain a prediction loss value;
updating the prediction model to be trained based on the prediction loss value to obtain a target prediction model;
and updating each updated feature extraction model based on the predicted loss value to obtain a target feature extraction model.
Optionally, the grouping module 100 is further configured to:
grouping the characteristics of the first sample based on preset service requirements to obtain a first sample characteristic group corresponding to the first sample and a plurality of single characteristics;
the prediction model determination module 400 is further configured to:
and inputting the first sample characteristic representation and the characteristic feature set corresponding to the monomer characteristic into a prediction model to be trained for model training to obtain the prediction loss value.
Optionally, the target prediction model is a fraud score prediction model; the first service scene comprises a credit scoring scene of a user, the second service scene comprises a fraud scoring scene of the user, the first sample comprises credit scoring data in a first participant, and the second sample is fraud scoring data in the first participant;
a model determination module 300 further configured to:
determining a first domain distinguishing loss value based on a first sample feature representation corresponding to credit scoring data in a first participant and a domain distinguishing model corresponding to the first sample feature representation, determining a second domain distinguishing loss value based on a second sample feature representation corresponding to fraud scoring data in the first participant and a domain distinguishing model corresponding to the fraud scoring data in the first participant, and updating each feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value;
the prediction model determination module 400 is further configured to:
and determining a prediction loss value and a fraud score prediction model based on a first sample feature representation corresponding to credit score data in a first participant and an interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
Optionally, the data processing apparatus based on transfer learning further includes:
grouping the characteristics of the samples to be predicted of the second service scene based on the preset service requirements to obtain a sample characteristic group to be predicted;
inputting a sample feature group to be predicted into a corresponding target feature extraction model to obtain a plurality of feature representations to be predicted;
and inputting the feature characterization set corresponding to the feature characterization to be predicted into the target prediction model to obtain a prediction result.
Optionally, the data processing apparatus based on transfer learning further includes:
grouping the characteristics of the samples to be predicted of the second service scene based on preset service requirements to obtain a sample characteristic group to be predicted and a plurality of monomer characteristics to be predicted;
and inputting the feature characterization to be predicted and the feature characterization set corresponding to the monomer feature to be predicted into the target prediction model to obtain a prediction result.
Furthermore, an embodiment of the present invention further provides a storage medium, where the storage medium stores a data processing program based on migratory learning, and the data processing program based on migratory learning implements the steps of the data processing method based on migratory learning as described above when executed by a processor.
The method implemented when the data processing program based on the migration learning running on the processor is executed may refer to each embodiment of the data processing method based on the migration learning of the present invention, and is not described herein again.
Furthermore, an embodiment of the present invention further provides a computer program product, which when executed, implements the steps of the data processing method based on the migration learning as described above.
The method implemented when the computer program product executes the data processing program based on the transfer learning may refer to each embodiment of the data processing method based on the transfer learning of the present invention, and is not described herein again.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) as described above and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present invention.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.

Claims (12)

1. A data processing method based on transfer learning is characterized by comprising the following steps:
grouping the characteristics of first samples of a first service scene based on preset service requirements to obtain a preset number of first sample characteristic groups, and grouping the characteristics of second samples of a second service scene based on the preset service requirements to obtain a preset number of second sample characteristic groups;
determining a first sample feature characterization based on each first sample feature group and the corresponding feature extraction model thereof, and determining a second sample feature characterization based on each second sample feature group and the corresponding feature extraction model thereof;
determining a first domain distinguishing loss value based on each first sample feature characterization and a corresponding domain distinguishing model thereof, determining a second domain distinguishing loss value based on each second sample feature characterization and a corresponding domain distinguishing model thereof, and updating each feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value;
and determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
2. The data processing method based on the transfer learning of claim 1, wherein the step of determining the first domain discrimination loss value based on each first sample feature characterization and its corresponding domain discrimination model comprises:
inputting each first sample characteristic representation into a corresponding domain distinguishing model to obtain a first domain distinguishing loss value corresponding to each first sample characteristic representation;
and updating the corresponding domain distinguishing model based on the first domain distinguishing loss values to obtain an updated domain distinguishing model, and updating the corresponding feature extraction model based on each first domain distinguishing loss value through domain confrontation learning to obtain an intermediate feature extraction model.
3. The data processing method based on the transfer learning of claim 2, wherein the step of determining a second domain discrimination loss value based on each second sample feature characterization and the corresponding domain discrimination model thereof, and updating each feature extraction model based on the first domain discrimination loss value and the second domain discrimination loss value comprises:
inputting each second sample characteristic representation into the corresponding updated domain distinguishing model to obtain a second domain distinguishing loss value corresponding to each second sample characteristic representation;
and updating the corresponding intermediate feature extraction model through domain confrontation learning based on each second domain distinguishing loss value to obtain an updated feature extraction model.
4. The data processing method based on the transfer learning of claim 1, wherein the step of determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model comprises:
inputting the characteristic feature set corresponding to the first sample characteristic representation into a prediction model to be trained for model training to obtain a prediction loss value;
updating the prediction model to be trained based on the prediction loss value to obtain a target prediction model;
and updating each updated feature extraction model based on the predicted loss value to obtain a target feature extraction model.
5. The data processing method based on transfer learning of claim 4, wherein the step of grouping the features of the first samples of the first service scenario based on the preset service requirement to obtain the preset number of first sample feature groups comprises:
grouping the characteristics of the first sample based on preset service requirements to obtain a first sample characteristic group corresponding to the first sample and a plurality of single characteristics;
the step of inputting the feature characteristic set corresponding to the first sample feature characterization into the prediction model to be trained for model training to obtain the prediction loss value comprises:
and inputting the first sample characteristic representation and the characteristic feature set corresponding to the monomer characteristic into a prediction model to be trained for model training to obtain the prediction loss value.
6. The migration learning-based data processing method according to claim 1, wherein the target prediction model is a fraud score prediction model; the first service scene comprises a credit scoring scene of a user, the second service scene comprises a fraud scoring scene of the user, the first sample comprises credit scoring data in a first participant, and the second sample is fraud scoring data in the first participant;
the step of determining a first domain discrimination loss value based on each first sample feature characterization and the corresponding domain discrimination model, determining a second domain discrimination loss value based on each second sample feature characterization and the corresponding domain discrimination model, and updating each feature extraction model based on the first domain discrimination loss value and the second domain discrimination loss value comprises:
determining a first domain distinguishing loss value based on a first sample feature representation corresponding to credit scoring data in a first participant and a domain distinguishing model corresponding to the first sample feature representation, determining a second domain distinguishing loss value based on a second sample feature representation corresponding to fraud scoring data in the first participant and a domain distinguishing model corresponding to the fraud scoring data in the first participant, and updating each feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value;
the step of determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model comprises the following steps:
and determining a prediction loss value and a fraud score prediction model based on a first sample feature representation corresponding to credit score data in a first participant and an interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
7. The migration learning-based data processing method according to any one of claims 1 to 6, wherein after the step of determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model, the migration learning-based data processing method further comprises:
grouping the characteristics of the samples to be predicted of the second service scene based on the preset service requirements to obtain a sample characteristic group to be predicted;
inputting a sample feature group to be predicted into a corresponding target feature extraction model to obtain a plurality of feature representations to be predicted;
and inputting the feature characterization set corresponding to the feature characterization to be predicted into the target prediction model to obtain a prediction result.
8. The data processing method based on the transfer learning of claim 7, wherein the step of grouping the features of the samples to be predicted of the second service scenario based on the preset service requirement to obtain the feature group of the samples to be predicted comprises:
grouping the characteristics of the samples to be predicted of the second service scene based on preset service requirements to obtain a sample characteristic group to be predicted and a plurality of monomer characteristics to be predicted;
the step of inputting the feature characterization set corresponding to the feature characterization to be predicted into the target prediction model to obtain the prediction result comprises the following steps:
and inputting the feature characterization to be predicted and the feature characterization set corresponding to the monomer feature to be predicted into the target prediction model to obtain a prediction result.
9. A data processing apparatus based on migration learning, characterized in that the data processing apparatus based on migration learning comprises:
the grouping module is used for grouping the characteristics of the first samples of the first service scene based on preset service requirements to obtain a preset number of first sample characteristic groups, and grouping the characteristics of the second samples of the second service scene based on the preset service requirements to obtain a preset number of second sample characteristic groups;
the characteristic extraction module is used for determining first sample characteristic characteristics based on each first sample characteristic group and the corresponding characteristic extraction model thereof, and determining second sample characteristic characteristics based on each second sample characteristic group and the corresponding characteristic extraction model thereof;
the model determining module is used for determining a first domain distinguishing loss value based on each first sample characteristic characterization and the corresponding domain distinguishing model thereof, determining a second domain distinguishing loss value based on each second sample characteristic characterization and the corresponding domain distinguishing model thereof, and updating each feature extraction model based on the first domain distinguishing loss value and the second domain distinguishing loss value;
and the prediction model determining module is used for determining a prediction loss value and a target prediction model based on the first sample feature characterization and the interpretable prediction model to be trained, and determining a target feature extraction model based on the prediction loss value and the updated feature extraction model.
10. A data processing apparatus based on migration learning, characterized in that the data processing apparatus based on migration learning comprises: memory, a processor and a migration learning based data processing program stored on the memory and executable on the processor, the migration learning based data processing program when executed by the processor implementing the steps of the migration learning based data processing method according to any one of claims 1 to 8.
11. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a migration learning based data processing program, which when executed by a processor implements the steps of the migration learning based data processing method according to any one of claims 1 to 8.
12. A computer program product, characterized in that it implements the steps of the migration learning based data processing method according to any one of claims 1 to 8 when executed.
CN202011393272.4A 2020-12-02 2020-12-02 Data processing method, device and equipment based on transfer learning and storage medium Pending CN112418443A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011393272.4A CN112418443A (en) 2020-12-02 2020-12-02 Data processing method, device and equipment based on transfer learning and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011393272.4A CN112418443A (en) 2020-12-02 2020-12-02 Data processing method, device and equipment based on transfer learning and storage medium

Publications (1)

Publication Number Publication Date
CN112418443A true CN112418443A (en) 2021-02-26

Family

ID=74829754

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011393272.4A Pending CN112418443A (en) 2020-12-02 2020-12-02 Data processing method, device and equipment based on transfer learning and storage medium

Country Status (1)

Country Link
CN (1) CN112418443A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112949752A (en) * 2021-03-25 2021-06-11 支付宝(杭州)信息技术有限公司 Training method and device of business prediction system
CN113378993A (en) * 2021-07-09 2021-09-10 深圳前海微众银行股份有限公司 Artificial intelligence based classification method, device, equipment and storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112949752A (en) * 2021-03-25 2021-06-11 支付宝(杭州)信息技术有限公司 Training method and device of business prediction system
CN113378993A (en) * 2021-07-09 2021-09-10 深圳前海微众银行股份有限公司 Artificial intelligence based classification method, device, equipment and storage medium

Similar Documents

Publication Publication Date Title
US11169827B2 (en) Resource loading at application startup using attributes of historical data groups
US11983646B2 (en) Bias scoring of machine learning project data
CN108280115B (en) Method and device for identifying user relationship
CN107103036B (en) Method and equipment for acquiring application downloading probability and programmable equipment
CN109471978B (en) Electronic resource recommendation method and device
US11748452B2 (en) Method for data processing by performing different non-linear combination processing
CN114036398B (en) Content recommendation and ranking model training method, device, equipment and storage medium
CN111950593A (en) Method and device for recommending model training
CN112418443A (en) Data processing method, device and equipment based on transfer learning and storage medium
CN110020022A (en) Data processing method, device, equipment and readable storage medium storing program for executing
CN110827120A (en) GAN network-based fuzzy recommendation method and device, electronic equipment and storage medium
CN112380104A (en) User attribute identification method and device, electronic equipment and storage medium
CN113886721B (en) Personalized interest point recommendation method and device, computer equipment and storage medium
CN112418442A (en) Data processing method, device, equipment and storage medium for federal transfer learning
CN112381236A (en) Data processing method, device, equipment and storage medium for federal transfer learning
CN115619448A (en) User loss prediction method and device, computer equipment and storage medium
CN112418441A (en) Data processing method, device and equipment based on transfer learning and storage medium
CN113869377A (en) Training method and device and electronic equipment
CN113378067A (en) Message recommendation method, device, medium, and program product based on user mining
CN111291868A (en) Network model training method, device, equipment and computer readable storage medium
CN114430504B (en) Recommendation method and related device for media content
CN114139052B (en) Ranking model training method for intelligent recommendation, intelligent recommendation method and device
CN111988407B (en) Content pushing method and related device
CN116205686A (en) Method, device, equipment and storage medium for recommending multimedia resources
CN112749214A (en) Updating method, device and medium of interactive content display mode and electronic equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination