CN110889759A - Credit data determination method, device and storage medium - Google Patents
Credit data determination method, device and storage medium Download PDFInfo
- Publication number
- CN110889759A CN110889759A CN201911147899.9A CN201911147899A CN110889759A CN 110889759 A CN110889759 A CN 110889759A CN 201911147899 A CN201911147899 A CN 201911147899A CN 110889759 A CN110889759 A CN 110889759A
- Authority
- CN
- China
- Prior art keywords
- credit
- data
- model
- comprehensive
- investigation
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The application discloses a method, a device and a storage medium for determining credit data, and belongs to the field of internet finance. The method comprises the following steps: acquiring first credit investigation characteristics corresponding to credit investigation data of a target user, wherein the credit investigation data comprises various types of credit investigation data; evaluating the first credit investigation characteristics through an FM model and a DNN model in the comprehensive credit evaluation model respectively to obtain a first credit score and a second credit score of the target user; and fusing the first credit score and the second credit score through an output layer of the comprehensive credit evaluation model to obtain a third credit score of the target user. According to the credit assessment method and the credit assessment device, the credit of the user can be assessed uniformly according to the incidence relation among various credit assessment data and the influence factors among the credit assessment data, so that the credit score obtained by assessment can represent the credit of the user more comprehensively and accurately, and the accuracy of determining the credit data is improved.
Description
Technical Field
The present application relates to the field of internet finance, and in particular, to a method, an apparatus, and a storage medium for determining credit data.
Background
In the field of internet finance, when a user applies for a financial product, a financial platform needs to acquire credit investigation data of the user, determine credit score capable of indicating credit degree of the user according to the credit investigation data of the user, and judge whether the user meets application conditions according to the credit score of the user.
At present, various types of credit assessment models have appeared, and different types of credit assessment models are used for assessing the credit of a user according to different types of credit assessment data, and the obtained credit scores are used for representing the credit of the user from different angles without any relation. When credit assessment is required to be performed on a user, various types of credit investigation data of the user can be obtained, feature extraction is performed on the various types of credit investigation data respectively to obtain credit investigation features of the various types of credit investigation data, then the credit investigation features of the various types of credit investigation data are used as input of credit assessment models of corresponding types respectively, the credit investigation features of the various types of credit investigation data are assessed and processed through the credit assessment models of corresponding types to obtain various credit scores corresponding to the various types of credit investigation data one by one, credit risks of the user are estimated according to the various credit scores, and whether the user meets application conditions is judged according to risk assessment results.
Disclosure of Invention
The embodiment of the application provides a method and a device for determining credit data and a storage medium, which can be used for solving the problem of low accuracy of determining the credit data in the related art. The technical scheme is as follows:
in one aspect, a method for determining credit data is provided, where the method includes:
acquiring first credit investigation characteristics corresponding to credit investigation data of a target user, wherein the credit investigation data comprises various types of credit investigation data;
taking the first credit investigation feature as an input of a comprehensive credit assessment model, and respectively assessing and processing the first credit investigation feature through an FM (frequency modulation) model and a DNN (digital noise network) model in the comprehensive credit assessment model to obtain a first credit score and a second credit score of the target user;
and fusing the first credit score and the second credit score through an output layer of the comprehensive credit evaluation model to obtain a third credit score of the target user.
Optionally, the obtaining of the first credit investigation feature of the credit investigation data of the target user includes:
performing feature extraction on credit investigation data of the target user to obtain second credit investigation features corresponding to the credit investigation data, wherein the second credit investigation features comprise credit investigation features corresponding to continuous credit investigation data and credit investigation features corresponding to discrete credit investigation data;
carrying out word embedding processing on the credit investigation characteristics corresponding to the discrete credit investigation data to obtain word embedding vectors corresponding to the discrete credit investigation data;
and splicing the credit investigation feature corresponding to the continuous credit investigation data and the word embedding vector corresponding to the discrete credit investigation data to obtain the first credit investigation feature.
Optionally, the multiple types of credit investigation data include behavior data on a preset application and credit investigation data on at least one credit investigation system.
Optionally, before the first credit investigation feature is used as an input of a comprehensive credit evaluation model, and the first credit investigation feature is evaluated by an FM model and a DNN model in the comprehensive credit evaluation model, the method further includes:
acquiring first credit investigation characteristics corresponding to credit investigation data of a plurality of sample users respectively and sample credit scores of the plurality of sample users;
training a comprehensive credit evaluation model to be trained according to first credit evaluation characteristics respectively corresponding to credit evaluation data of a plurality of sample users and sample credit scores of the plurality of sample users to obtain the comprehensive credit evaluation model; the comprehensive credit evaluation model to be trained comprises an FM model to be trained, a DNN model to be trained and an output layer, wherein input data of the FM model to be trained and input data of the DNN model to be trained are the same, and the output layer is used for carrying out fusion processing on output data of the FM model to be trained and output data of the DNN model to be trained to obtain output data of the comprehensive credit evaluation model to be trained.
Optionally, the training the comprehensive credit evaluation model to be trained according to the first credit investigation features respectively corresponding to the credit investigation data of the plurality of sample users and the sample credit scores of the plurality of sample users includes:
taking first credit investigation characteristics corresponding to credit investigation data of the plurality of sample users as input of the comprehensive credit assessment model to be trained, and determining predicted credit scores of the plurality of sample users through the comprehensive credit assessment model to be trained;
determining a comprehensive prediction error of the comprehensive credit evaluation model to be trained according to the prediction credit scores of the plurality of sample users and the sample credit scores;
according to a back propagation algorithm, performing back propagation on the comprehensive prediction error so as to update the model parameters of the comprehensive credit evaluation model to be trained;
and determining the comprehensive credit evaluation model to be trained after the model parameters are updated as the comprehensive credit evaluation model.
Optionally, the determining, by the to-be-trained comprehensive credit evaluation model, the predicted credit scores of the plurality of sample users includes:
evaluating first credit investigation characteristics respectively corresponding to credit investigation data of the plurality of sample users through an FM model to be trained and a DNN model to be trained in the comprehensive credit evaluation model to be trained to obtain first prediction credit scores and second prediction credit scores of the plurality of sample users;
and performing fusion processing on the first prediction credit scores and the second prediction credit scores of the plurality of sample users through an output layer in the comprehensive credit evaluation model to be trained to obtain the prediction credit scores of the plurality of sample users.
Optionally, the sample credit scores of the plurality of sample users include a plurality of credit scores corresponding to the plurality of types of credit data one to one;
determining a composite prediction error based on the prediction credit scores of the plurality of sample users and the sample credit scores, comprising:
determining a prediction error corresponding to each credit score according to the prediction credit scores of the plurality of sample users and each credit score in the plurality of credit scores to obtain a plurality of prediction errors;
determining weights for the plurality of prediction errors;
determining the composite prediction error based on the plurality of prediction errors and the weight of each prediction error.
In another aspect, an apparatus for determining credit data is provided, the apparatus including:
the first acquisition module is used for acquiring first credit investigation characteristics of credit investigation data of a target user, wherein the credit investigation data comprises various types of credit investigation data;
the evaluation module is used for taking the first credit investigation feature as an input of a comprehensive credit evaluation model, and evaluating the first credit investigation feature through an FM model and a DNN model in the comprehensive credit evaluation model respectively to obtain a first credit score and a second credit score of the target user;
and the processing module is used for carrying out fusion processing on the first credit score and the second credit score through an output layer of the comprehensive credit evaluation model to obtain a third credit score of the target user.
Optionally, the obtaining module is configured to:
performing feature extraction on credit investigation data of the target user to obtain second credit investigation features corresponding to the credit investigation data, wherein the second credit investigation features comprise credit investigation features corresponding to continuous credit investigation data and credit investigation features corresponding to discrete credit investigation data;
carrying out word embedding processing on the credit investigation characteristics corresponding to the discrete credit investigation data to obtain word embedding vectors corresponding to the discrete credit investigation data;
and splicing the credit investigation feature corresponding to the continuous credit investigation data and the word embedding vector corresponding to the discrete credit investigation data to obtain the first credit investigation feature.
Optionally, the multiple types of credit investigation data include behavior data on a preset application and credit investigation data on at least one credit investigation system.
Optionally, the apparatus further comprises:
the second acquisition module is used for acquiring first credit investigation characteristics corresponding to credit investigation data of a plurality of sample users respectively and sample credit scores of the plurality of sample users;
the training module is used for training a comprehensive credit evaluation model to be trained according to first credit evaluation characteristics corresponding to credit evaluation data of a plurality of sample users and sample credit scores of the plurality of sample users to obtain the comprehensive credit evaluation model; the comprehensive credit evaluation model to be trained comprises an FM model to be trained, a DNN model to be trained and an output layer, wherein input data of the FM model to be trained and input data of the DNN model to be trained are the same, and the output layer is used for carrying out fusion processing on output data of the FM model to be trained and output data of the DNN model to be trained to obtain output data of the comprehensive credit evaluation model to be trained.
Optionally, the training module comprises:
the first determining unit is used for taking first credit investigation features respectively corresponding to credit investigation data of the plurality of sample users as input of the comprehensive credit assessment model to be trained, and determining the predicted credit scores of the plurality of sample users through the comprehensive credit assessment model to be trained;
the second determining unit is used for determining the comprehensive prediction error of the comprehensive credit evaluation model to be trained according to the prediction credit scores of the plurality of sample users and the sample credit scores;
the propagation unit is used for performing back propagation on the comprehensive prediction error according to a back propagation algorithm so as to update the model parameters of the comprehensive credit evaluation model to be trained;
and the third determining unit is used for determining the comprehensive credit evaluation model to be trained after the model parameters are updated as the comprehensive credit evaluation model.
Optionally, the first determining unit is configured to:
evaluating first credit investigation characteristics respectively corresponding to credit investigation data of the plurality of sample users through an FM model to be trained and a DNN model to be trained in the comprehensive credit evaluation model to be trained to obtain first prediction credit scores and second prediction credit scores of the plurality of sample users;
and performing fusion processing on the first prediction credit scores and the second prediction credit scores of the plurality of sample users through an output layer in the comprehensive credit evaluation model to be trained to obtain the prediction credit scores of the plurality of sample users.
Optionally, the second determining unit is configured to;
determining a composite prediction error based on the prediction credit scores of the plurality of sample users and the sample credit scores, comprising:
determining a prediction error corresponding to each credit score according to the prediction credit scores of the plurality of sample users and each credit score in the plurality of credit scores to obtain a plurality of prediction errors;
determining weights for the plurality of prediction errors;
determining the composite prediction error based on the plurality of prediction errors and the weight of each prediction error.
In another aspect, an apparatus for determining credit data is provided, the apparatus including:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to perform the steps of any of the above-described methods of credit data determination.
In another aspect, a computer-readable storage medium is provided, having instructions stored thereon, which when executed by a processor, implement the steps of any one of the above-described methods for determining credit data.
In another aspect, a computer program product is provided, which when executed, is configured to implement the steps of any one of the above-mentioned methods for determining credit data.
The technical scheme provided by the embodiment of the application has the following beneficial effects:
in the embodiment of the application, credit investigation characteristics of various types of credit investigation data are input into a comprehensive credit assessment model for assessment processing, so that the comprehensive credit assessment model can uniformly assess the credit of the user according to the incidence relation and the influence factors among various types of credit investigation data, the credit score obtained by assessment can represent the credit of the user more comprehensively and accurately, and the accuracy of determining the credit data is improved. In addition, the comprehensive credit evaluation model comprises an FM model and a DNN model, low-order features and high-order features of credit investigation data can be extracted through the two models respectively, the low-order features and the high-order features are evaluated respectively, and the two obtained credit scores are fused to obtain a credit score which is finally used for representing the credit of the user.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flowchart of a training method for a comprehensive credit assessment model according to an embodiment of the present disclosure;
fig. 2 is a flowchart of a method for determining credit data according to an embodiment of the present application;
FIG. 3 is a block diagram of a comprehensive credit evaluation model according to an embodiment of the present disclosure;
fig. 4 is a block diagram of a device for determining credit data according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of a credit data determination apparatus according to an embodiment of the present application.
Detailed Description
To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.
Before explaining the embodiments of the present application in detail, an application scenario of the embodiments of the present application will be described. The embodiment of the application is applied to an internet financial scene and used for performing credit evaluation on the user in the internet financial scene.
In the related art, when an application platform evaluates credit of a user aiming at a financial scene, behavior data of the user on application can be acquired, and the behavior data is evaluated and processed through a credit evaluation model built by the application platform according to the behavior data to obtain application credit score. Then, a first credit score obtained by the first credit investigation system performing credit assessment on the user is obtained through a system interface of the first credit investigation system, a second credit score obtained by the second credit investigation system performing credit assessment on the user is obtained through a system interface of the second credit investigation system, and the credit risk of the user is estimated by combining the application credit score, the first credit score and the second credit score.
The application may be any application, such as an e-commerce application or an information aggregation application. The first credit investigation system can be a credit investigation system of an authority organization, credit investigation data of the first credit investigation system mainly comes from various large financial institutions, social security of public deposit, telecommunication and the like, the authority is strong, the data is complete, and the first credit investigation system can be used for evaluating personal assets, putting credit in banks, credit card lines and the like. For example, the first credit investigation system is a Chinese people bank credit investigation system, and the first credit score is a number interpretation score. The second credit investigation system can be a network credit investigation system, the credit investigation data of the network credit investigation system mainly comes from an internet platform, and the credit investigation data is captured or acquired by interface cooperation by using the internet technology. For example, the second credit system is the peer shield or the Bai Jie credit system, and the second credit score is the peer shield score or the Bai Jie score.
However, when the credit assessment problem is divided into separate and independent tasks for assessment, the rich association relationship among the tasks is ignored, the generalization capability is poor, and the credit scores of the separate assessments are not associated with each other and are not comparable, so that it is difficult to integrate the credit scores to assess the credit risk of the user accurately. For example, when the application credit score and the first credit score are high and the second credit score is low, it is difficult to accurately estimate the credit risk of the user in such a case.
In the embodiment of the application, in order to solve the problems that the generalization capability is poor, the credit evaluation cannot be accurately performed and the accuracy of determining the credit data is low due to the fact that the credit evaluation problem is decomposed into separate and independent tasks to be evaluated respectively in the related art, a method capable of sharing the correlation information among multiple tasks and comprehensively performing the credit evaluation on a user, namely a method for determining the credit data of the multiple tasks is provided, so that the generalization capability and the accuracy of the credit evaluation are improved.
Next, an implementation environment related to the embodiments of the present application will be described.
The method for determining the credit data provided by the embodiment of the application can be applied to electronic equipment, and the electronic equipment can be a terminal or a server. For example, the electronic device may be a background server of a designated application or a background server of a credit investigation system, which is not limited in this embodiment of the present application. The terminal can be a mobile phone, a tablet computer, a computer or the like. Of course, the method may also be applied to other electronic devices according to actual needs, and this is not limited in this embodiment of the application.
It should be noted that, in order to improve the accuracy of determining credit data, a comprehensive credit evaluation model capable of comprehensively evaluating multiple types of credit investigation data is constructed in the embodiments of the present application, and the comprehensive credit evaluation model is composed of two credit evaluation models. Before the credit evaluation is performed by the comprehensive credit evaluation model, the comprehensive credit evaluation model needs to be trained based on sample data, and then the training process of the comprehensive credit evaluation model is described in detail.
Fig. 1 is a flowchart of a training method of a comprehensive credit evaluation model according to an embodiment of the present application, and as shown in fig. 1, the method includes the following steps:
step 101: and acquiring credit investigation data and sample credit scores of a plurality of sample users, wherein the credit investigation data comprises a plurality of types of credit investigation data.
The sample user may be a user who has applied for a financial product in history and meets training requirements. The credit investigation data for each sample user may include multiple types of credit investigation data, and each type of credit investigation data may be used separately for credit assessment to derive a corresponding type of credit score. For example, the credit data types can be task data corresponding to the existing credit evaluation models.
As an example, the plurality of types of credit investigation data comprise behavior data on a preset application and credit investigation data on at least one credit investigation system.
The preset application may be any application, such as an e-commerce application or an information aggregation application. The at least one credit investigation system may comprise at least one first credit investigation system and at least one second credit investigation system.
The first credit investigation system can be a credit investigation system of an authority organization, credit investigation data of the first credit investigation system mainly comes from various large financial institutions, social security of public deposit, telecommunication and the like, the authority is strong, the data is complete, and the first credit investigation system can be used for evaluating personal assets, putting credit in banks, credit card lines and the like. For example, the first credit investigation system is a Chinese people bank credit investigation system, and the credit score corresponding to the first credit investigation system is a pedestrian digit interpretation score.
The second credit investigation system can be a network credit investigation system, the credit investigation data of the network credit investigation system mainly comes from an internet platform, and the credit investigation data is captured or acquired by interface cooperation by using the internet technology. For example, the second credit investigation system is the peer shield or the Bai Rou credit investigation system, and the credit score corresponding to the second credit investigation system is the peer shield score or the Bai Rou score.
As one example, the sample credit score for each sample user may include a plurality of credit scores, one for one, corresponding to a plurality of types of credit data. For example, if the plurality of types of credit investigation data include behavior data on a preset application and credit investigation data on at least one credit investigation system, the sample credit score includes an application credit score corresponding to the behavior data on the preset application and a credit score corresponding to the credit investigation data on each credit investigation system.
For example, if the plurality of types of credit investigation data of the sample user include behavior data of e-commerce application, credit investigation data of a chinese people bank credit investigation system, a peer-to-peer credit investigation system, and credit investigation data of a Bai Rou credit investigation system, the sample credit score of the sample user may include e-commerce application credit score, a pedestrian digital interpretation score, a peer-to-peer score, and a Bai Rou score.
Step 102: and performing feature extraction on the credit investigation data of each sample user to determine a first credit investigation feature corresponding to the credit investigation data of each sample user.
As an example, for each sample user of the plurality of sample users, feature extraction may be performed on credit data of the sample user to obtain a second credit feature corresponding to the credit data, where the credit data of the sample user may include continuous credit data and discrete credit data, and correspondingly, the second credit feature may include a credit feature corresponding to the continuous credit data and a credit feature corresponding to the discrete credit data. And then, carrying out word embedding processing on the credit investigation features corresponding to the discrete credit investigation data to obtain word embedding vectors corresponding to the discrete credit investigation data, and splicing the credit investigation features corresponding to the continuous credit investigation data and the word embedding vectors corresponding to the discrete credit investigation data to obtain first credit investigation features corresponding to the credit investigation data of each sample user.
As an example, the word embedding layer may perform word embedding processing on the credit investigation features corresponding to the discrete credit investigation data to map the discrete credit investigation features corresponding to the discrete credit investigation data to a low-dimensional vector. By way of example, the word embedding layer may be a locally connected like structure.
Step 103: and training the comprehensive credit evaluation model to be trained according to the first credit evaluation characteristics corresponding to the credit evaluation data of the sample users and the sample credit scores of the sample users to obtain the comprehensive credit evaluation model.
The comprehensive credit evaluation model to be trained comprises an FM (factor decomposition Machine) model to be trained, a DNN (Deep Neural Networks) model to be trained and an output layer.
The input data of the FM model to be trained and the input data of the DNN model to be trained are the same, and are the first credit investigation characteristics corresponding to the credit investigation data of the plurality of sample users respectively. The output layer is used for carrying out fusion processing on the output data of the FM model to be trained and the output data of the DNN model to be trained to obtain the output data of the comprehensive credit evaluation model to be trained.
It should be noted that the FM model to be trained can learn the low-order features of the credit investigation data, and the DNN model to be trained can learn the high-order features of the credit investigation data, so that the comprehensive credit assessment model to be trained can learn the low-order features and the high-order features of various types of credit investigation data through the FM model to be trained and the DNN model to be trained by combining the advantages of the breadth and depth models, and further perform the assessment processing on various types of credit investigation data by combining the low-order features and the high-order features of various types of credit investigation data.
In addition, the comprehensive credit evaluation model to be trained is an end-to-end model, and feature engineering is not needed. And the FM model to be trained and the DNN model to be trained share the same input, so that the training is more efficient.
As an example, the output layer may be a sigmoid (S-shaped growth curve) function, and a calculation formula for performing fusion processing on the output data of the FM model to be trained and the output data of the DNN model to be trained through the sigmoid function is as follows:
y＝sigmoid(y1+y2) (1)
wherein y is the output data of the comprehensive credit evaluation model to be trained, y1For the output data of the FM model to be trained, y2Is the output data of the DNN model to be trained.
As an example, according to the first credit investigation characteristics corresponding to the credit investigation data of the plurality of sample users, respectively, and the sample credit scores of the plurality of sample users, the operation of training the comprehensive credit evaluation model to be trained may include the following steps:
1) and taking the first credit investigation characteristics corresponding to the credit investigation data of the plurality of sample users as the input of the comprehensive credit assessment model to be trained, and determining the predicted credit scores of the plurality of sample users through the comprehensive credit assessment model to be trained.
As an example, the FM model to be trained and the DNN model to be trained in the comprehensive credit evaluation model to be trained may be used to respectively evaluate first credit evaluation features corresponding to credit evaluation data of a plurality of sample users, so as to obtain first predicted credit scores and second predicted credit scores of the plurality of sample users, and the first predicted credit scores and the second predicted credit scores of the plurality of sample users may be fused through an output layer in the comprehensive credit evaluation model to be trained, so as to obtain predicted credit scores of the plurality of sample users.
As one example, the FM model to be trained may determine the first predicted credit score by the following equation:
wherein, yFMIs the output data of the FM, i.e., the first predicted credit score; n represents a feature dimension of the first credit investigation feature; x is the number ofiThe ith characteristic in the first credit investigation characteristic is represented; x is the number ofjRepresenting the jth characteristic in the first credit characteristic; x is the number ofixjRepresenting a combined feature; w is aijRepresenting the importance of the combined features for the combined parameters; w is aiThe single characteristic parameter represents the importance of the single characteristic; w is a0Is a preset parameter.
As an example, if the first predicted credit score of the FM output to be trained is yFMSecond predictive Credit score of DNN output to be trained as yDNNAnd the output layer is a sigmoid function, and the predicted credit score y of the sample user is sigmoid (y)FM+yDNN)。
As an example, the DNN model to be trained functions to construct high-dimensional features, and the input of the DNN model is also the first credit feature, i.e. weight sharing with the FM model to be trained.
As an example, the input processing manner in the DNN model to be trained may adopt forward propagation, and the input of the DNN model to be trained is assumed to be α(0)＝(e1,e2,...,en) Then α(0)As an input to the next DNN hidden layer, the feed forward process may be α(l+1)＝σ(W(l)α(l)+b(l))。
As an example, the characteristic length of the first credit characteristic may be set to a specified length, such as a specified length of 64. In addition, the DNN model to be trained may employ n fully-connected layers, where n is an integer greater than 1, for example, n is 3.
As an example, a dropout coefficient may be set for each layer network layer of the FM model to be trained and the DNN model to be trained. For example, the dropout coefficient of the FM model to be trained is a first coefficient, such as 0.85; the dropout coefficient of the DNN model to be trained is a second coefficient, for example, 0.9.
2) And determining a comprehensive prediction error of the comprehensive credit evaluation model to be trained according to the prediction credit scores of the plurality of sample users and the sample credit scores.
That is, the predicted credit scores for a plurality of sample users may be compared to the sample credit scores to determine a composite prediction error for the composite credit evaluation model to be trained. As one example, the composite prediction error of the composite credit evaluation model to be trained may be represented by a loss function.
As an example, the sample credit scores of the plurality of sample users include a plurality of credit scores corresponding to a plurality of types of credit investigation data one-to-one, and when determining the comprehensive prediction error of the comprehensive credit evaluation model to be trained, a prediction error corresponding to each credit score may be determined according to the prediction credit scores of the plurality of sample users and each credit score of the plurality of credit scores, so as to obtain a plurality of prediction errors, and the comprehensive prediction error may be determined according to the plurality of prediction errors.
As an example, determining the composite prediction error from the plurality of prediction errors includes the following implementations:
in a first implementation, multiple prediction errors are used as the combined prediction error.
In a second implementation, the multiple prediction errors are accumulated to obtain a comprehensive prediction error.
In a third implementation, the weights of the plurality of prediction errors are determined, and the comprehensive prediction error is determined according to the plurality of prediction errors and the weight of each prediction error.
In a third implementation manner, in order to meet the scene requirements of different financial scenes, corresponding weights may be set for multiple prediction errors, so that the learning objectives under different scenes are adjusted by adjusting the weights of the multiple prediction errors, and different emphasis on different products or people is achieved. For example, if a scene with a higher emphasis on long-term risk is used, the weight of the prediction error corresponding to credit data for evaluating long-term credit may be set to be larger.
As an example, each prediction error may be multiplied by a corresponding weight, and then the products of the various prediction errors and the corresponding weights may be summed to obtain a composite prediction error.
As an example, the plurality of prediction errors may be represented by loss functions, and the loss function of the comprehensive credit evaluation model to be trained may be determined according to the plurality of loss functions and the weight of each loss function. The weights of the various loss functions can be custom set.
Illustratively, if the plurality of loss functions includes 3 loss functions, each loss function is less1、loss2And loss3The weights of these 3 kinds of loss functions are a1、a2And a3Then, the loss function loss of the integrated credit evaluation model to be trained is a1*loss1+a2*loss2+a3*loss3。
3) And according to a back propagation algorithm, performing back propagation on the comprehensive prediction error so as to update the model parameters of the comprehensive credit evaluation model to be trained.
The back propagation algorithm may be a gradient descent method, or other back propagation algorithms.
4) And determining the comprehensive credit evaluation model to be trained after the model parameters are updated as the trained comprehensive credit evaluation model.
In the embodiment of the application, in the process of training the comprehensive credit evaluation model to be trained, an alternative training or joint training mode can be adopted for training. In addition, a main task and an auxiliary task can be respectively set for various types of credit investigation data, different training probabilities are respectively set for the main task and the auxiliary task, the training probability of the main task is greater than that of the auxiliary task, for example, the training probability of the main task is 0.6, and the training probability of the auxiliary task is 0.2, so that the main task can be fully trained under limited training rounds.
As an example, the to-be-trained comprehensive credit evaluation model can be trained by using an Adam optimizer, for example, the initial learning rate is set to be 0.015, the learning rate is attenuated once every 2000 training steps, and the attenuation coefficient is 0.95.
In some examples, first credit investigation features and sample credit scores corresponding to credit investigation data of sample users in different areas may also be obtained, and according to the first credit investigation features and the sample credit scores corresponding to the credit investigation data of the sample users in each area, the comprehensive credit evaluation model to be trained is trained to obtain a comprehensive credit evaluation model corresponding to the area. The comprehensive credit evaluation model corresponding to each area is used for carrying out credit evaluation on the users of the area.
Wherein the area may be province or city, etc. By learning the internal characteristics of the credit investigation data of the sample users in different areas, the effect of differential estimation of the users in different areas by using a single model can be realized, and compared with the traditional machine learning method, the effect of differential estimation in different areas can be realized without special processing on the characteristics.
After the comprehensive credit assessment model is trained, the credit assessment can be performed on the user based on the trained comprehensive credit assessment model to determine credit data capable of representing the credit degree of the user. Fig. 2 is a flowchart of a method for determining credit data according to an embodiment of the present application, where the method is applied to a device for determining credit data, where the device for determining credit data may be a terminal or a server, as shown in fig. 2, and the method includes the following steps:
step 201: the method comprises the steps of obtaining a first credit investigation characteristic corresponding to credit investigation data of a target user, wherein the credit investigation data comprises various types of credit investigation data.
Wherein the target user is a user to be credit evaluated. The multiple types of credit investigation data include behavior data on a preset application and credit investigation data on at least one credit investigation system, and the detailed description of the multiple types of credit investigation data may refer to the relevant identifier in step 101, which is not described herein again in this embodiment of the present application.
As an example, the operation of acquiring the first credit investigation feature corresponding to the credit investigation data of the target user includes: carrying out feature extraction on credit investigation data of a target user to obtain second credit investigation features corresponding to the credit investigation data, wherein the second credit investigation features comprise credit investigation features corresponding to continuous credit investigation data and credit investigation features corresponding to discrete credit investigation data; carrying out word embedding processing on the credit investigation characteristics corresponding to the discrete credit investigation data to obtain word embedding vectors corresponding to the discrete credit investigation data; and splicing the credit investigation feature corresponding to the continuous credit investigation data and the word embedded vector corresponding to the discrete credit investigation data to obtain a first credit investigation feature.
Step 202: and taking the first credit investigation characteristic as an input of a comprehensive credit assessment model, and respectively assessing and processing the first credit investigation characteristic through an FM (frequency modulation) model and a DNN (digital noise network) model in the comprehensive credit assessment model to obtain a first credit score and a second credit score of the target user.
Wherein the comprehensive credit evaluation model comprises an FM model, a DNN model and an output layer. The input data of the FM model and the input data of the DNN model are the same and are first credit investigation characteristics corresponding to credit investigation data of the target user. And the output layer is used for carrying out fusion processing on the output data of the FM model and the output data of the DNN model to obtain the output data of the comprehensive credit evaluation model.
The FM model can learn the low-order characteristics of the credit investigation data, the DNN model can learn the high-order characteristics of the credit investigation data, and therefore the comprehensive credit evaluation model can extract the low-order characteristics and the high-order characteristics of various types of credit investigation data through the FM model and the DNN model by combining the advantages of the breadth model and the depth model, and further evaluate and process various types of credit investigation data in a unified mode by combining the low-order characteristics and the high-order characteristics of various types of credit investigation data.
Step 203: and fusing the first credit score and the second credit score through an output layer of the comprehensive credit evaluation model to obtain a third credit score of the target user.
As an example, the output layer may be a sigmoid function, and a calculation formula for performing fusion processing on the first credit score and the second credit score through the sigmoid function is as follows:
y3＝sigmoid(y1+y2)
wherein, y3Score for third Credit, y1Score first credit, y2Scoring a second credit.
Referring to fig. 3, fig. 3 is a model structure diagram of a comprehensive credit evaluation model according to an embodiment of the present application, as shown in fig. 3, the comprehensive credit evaluation model includes an FM model, a DNN model, and a sigmoid layer. The input data of the FM model and the input data of the DNN model are the same and are first credit investigation characteristics x corresponding to credit investigation data of the target user, and the first credit investigation characteristics are evaluated through the FM model to obtain a first credit score y1Evaluating the first credit investigation characteristics through a DNN model to obtain a second credit score y2And scoring the first credit y through the sigmoid layer1And a second credit score y2Performing fusion processing to obtain a third credit score y3。
In the embodiment of the application, credit investigation characteristics of various types of credit investigation data are input into a comprehensive credit assessment model for assessment processing, so that the comprehensive credit assessment model can uniformly assess the credit of the user according to the incidence relation and the influence factors among various types of credit investigation data, the credit score obtained by assessment can represent the credit of the user more comprehensively and accurately, and the accuracy of determining the credit data is improved. In addition, the comprehensive credit evaluation model comprises an FM model and a DNN model, low-order features and high-order features of credit investigation data can be extracted through the two models respectively, the low-order features and the high-order features are evaluated respectively, and the two obtained credit scores are fused to obtain a credit score which is finally used for representing the credit of the user.
Fig. 4 is a block diagram of an apparatus for determining credit data according to an embodiment of the present application, where as shown in fig. 4, the apparatus includes: a first acquisition module 401, an evaluation module 402 and a processing module 403.
A first obtaining module 401, configured to obtain a first credit investigation feature of credit investigation data of a target user, where the credit investigation data includes multiple types of credit investigation data;
an evaluation module 402, configured to use the first credit investigation feature as an input of a comprehensive credit evaluation model, and evaluate the first credit investigation feature through an FM model and a DNN model in the comprehensive credit evaluation model respectively to obtain a first credit score and a second credit score of the target user;
and a processing module 403, configured to perform fusion processing on the first credit score and the second credit score through an output layer of the comprehensive credit evaluation model to obtain a third credit score of the target user.
Optionally, the obtaining module 401 is configured to:
performing feature extraction on credit investigation data of the target user to obtain second credit investigation features corresponding to the credit investigation data, wherein the second credit investigation features comprise credit investigation features corresponding to continuous credit investigation data and credit investigation features corresponding to discrete credit investigation data;
carrying out word embedding processing on the credit investigation characteristics corresponding to the discrete credit investigation data to obtain word embedding vectors corresponding to the discrete credit investigation data;
and splicing the credit investigation feature corresponding to the continuous credit investigation data and the word embedding vector corresponding to the discrete credit investigation data to obtain the first credit investigation feature.
Optionally, the multiple types of credit investigation data include behavior data on a preset application and credit investigation data on at least one credit investigation system.
Optionally, the apparatus further comprises:
the second acquisition module is used for acquiring first credit investigation characteristics corresponding to credit investigation data of a plurality of sample users respectively and sample credit scores of the plurality of sample users;
the training module is used for training the comprehensive credit evaluation model to be trained according to the first credit evaluation characteristics corresponding to the credit evaluation data of the plurality of sample users and the sample credit scores of the plurality of sample users to obtain the comprehensive credit evaluation model; the comprehensive credit evaluation model to be trained comprises an FM model to be trained, a DNN model to be trained and an output layer, wherein the input data of the FM model to be trained and the input data of the DNN model to be trained are the same, and the output layer is used for carrying out fusion processing on the output data of the FM model to be trained and the output data of the DNN model to be trained to obtain the output data of the comprehensive credit evaluation model to be trained.
Optionally, the training module comprises:
the first determining unit is used for taking first credit investigation features respectively corresponding to credit investigation data of the plurality of sample users as input of the comprehensive credit assessment model to be trained, and determining the predicted credit scores of the plurality of sample users through the comprehensive credit assessment model to be trained;
the second determining unit is used for determining the comprehensive prediction error of the comprehensive credit evaluation model to be trained according to the prediction credit scores of the plurality of sample users and the sample credit scores;
the propagation unit is used for performing back propagation on the comprehensive prediction error according to a back propagation algorithm so as to update the model parameters of the comprehensive credit evaluation model to be trained;
and the third determining unit is used for determining the comprehensive credit evaluation model to be trained after the model parameters are updated as the comprehensive credit evaluation model.
Optionally, the first determining unit is configured to:
evaluating first credit investigation characteristics respectively corresponding to credit investigation data of the plurality of sample users through an FM model to be trained and a DNN model to be trained in the comprehensive credit evaluation model to be trained to obtain first prediction credit scores and second prediction credit scores of the plurality of sample users;
and performing fusion processing on the first prediction credit scores and the second prediction credit scores of the plurality of sample users through an output layer in the comprehensive credit evaluation model to be trained to obtain the prediction credit scores of the plurality of sample users.
Optionally, the second determining unit is configured to;
determining a composite prediction error based on the prediction credit scores of the plurality of sample users and the sample credit scores, comprising:
determining a prediction error corresponding to each credit score according to the prediction credit scores of the plurality of sample users and each credit score in the plurality of credit scores to obtain a plurality of prediction errors;
determining weights for the plurality of prediction errors;
the composite prediction error is determined based on the plurality of prediction errors and the weight of each prediction error.
In the embodiment of the application, credit investigation characteristics of various types of credit investigation data are input into a comprehensive credit assessment model for assessment processing, so that the comprehensive credit assessment model can uniformly assess the credit of the user according to the incidence relation and the influence factors among various types of credit investigation data, the credit score obtained by assessment can represent the credit of the user more comprehensively and accurately, and the accuracy of determining the credit data is improved. In addition, the comprehensive credit evaluation model comprises an FM model and a DNN model, low-order features and high-order features of credit investigation data can be extracted through the two models respectively, the low-order features and the high-order features are evaluated respectively, and the two obtained credit scores are fused to obtain a credit score which is finally used for representing the credit of the user.
It should be noted that: in the above embodiment, when performing credit evaluation, the determination apparatus for credit data is only illustrated by dividing the functional modules, and in practical applications, the function distribution may be completed by different functional modules according to needs, that is, the internal structure of the apparatus is divided into different functional modules to complete all or part of the functions described above. In addition, the credit data determination apparatus provided in the above embodiment and the credit data determination method embodiment belong to the same concept, and specific implementation processes thereof are described in the method embodiment and are not described herein again.
Fig. 5 is a schematic structural diagram of a device 500 for determining credit data according to an embodiment of the present application, where the device 500 for determining credit data may be an electronic device such as a terminal or a server, and the device 500 for determining credit data may generate a relatively large difference due to different configurations or performances, and may include one or more processors (CPUs) 501 and one or more memories 502, where the memory 502 stores at least one instruction, and the at least one instruction is loaded and executed by the processor 501 to implement the method for determining credit data according to the above-mentioned method embodiments. Certainly, the credit data determination apparatus 500 may further include a wired or wireless network interface, a keyboard, an input/output interface, and other components to facilitate input and output, and the credit data determination apparatus 500 may further include other components for implementing functions of the device, which is not described herein again.
In an exemplary embodiment, a computer-readable storage medium is also provided, which has instructions stored thereon, which when executed by a processor, implement the above-described credit data determination method.
In an exemplary embodiment, a computer program product is also provided for implementing the above-described method of determining credit data when executed.
It should be understood that reference to "a plurality" herein means two or more. "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The above description is only exemplary of the present application and should not be taken as limiting, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (10)
1. A method for determining credit data, the method comprising:
acquiring first credit investigation characteristics corresponding to credit investigation data of a target user, wherein the credit investigation data comprises various types of credit investigation data;
taking the first credit investigation feature as an input of a comprehensive credit assessment model, and respectively assessing and processing the first credit investigation feature through a Factorization Machine (FM) model and a Deep Neural Network (DNN) model in the comprehensive credit assessment model to obtain a first credit score and a second credit score of the target user;
and fusing the first credit score and the second credit score through an output layer of the comprehensive credit evaluation model to obtain a third credit score of the target user.
2. The method of claim 1, wherein the obtaining of the first credit investigation feature corresponding to the credit investigation data of the target user comprises:
performing feature extraction on credit investigation data of the target user to obtain second credit investigation features corresponding to the credit investigation data, wherein the second credit investigation features comprise credit investigation features corresponding to continuous credit investigation data and credit investigation features corresponding to discrete credit investigation data;
carrying out word embedding processing on the credit investigation characteristics corresponding to the discrete credit investigation data to obtain word embedding vectors corresponding to the discrete credit investigation data;
and splicing the credit investigation feature corresponding to the continuous credit investigation data and the word embedding vector corresponding to the discrete credit investigation data to obtain the first credit investigation feature.
3. The method of claim 1, wherein the plurality of types of credit investigation data comprise behavior data on a preset application and credit investigation data on at least one credit investigation system.
4. The method according to any one of claims 1-3, wherein the inputting the first credit investigation feature as an input of a comprehensive credit assessment model, before the first credit investigation feature is assessed and processed by an FM model and a DNN model of the comprehensive credit assessment model, further comprises:
acquiring first credit investigation characteristics corresponding to credit investigation data of a plurality of sample users respectively and sample credit scores of the plurality of sample users;
training a comprehensive credit evaluation model to be trained according to first credit evaluation characteristics respectively corresponding to credit evaluation data of a plurality of sample users and sample credit scores of the plurality of sample users to obtain the comprehensive credit evaluation model; the comprehensive credit evaluation model to be trained comprises an FM model to be trained, a DNN model to be trained and an output layer, wherein input data of the FM model to be trained and input data of the DNN model to be trained are the same, and the output layer is used for carrying out fusion processing on output data of the FM model to be trained and output data of the DNN model to be trained to obtain output data of the comprehensive credit evaluation model to be trained.
5. The method according to claim 4, wherein the training of the comprehensive credit evaluation model to be trained according to the first credit investigation features respectively corresponding to the credit investigation data of the plurality of sample users and the sample credit scores of the plurality of sample users comprises:
taking first credit investigation characteristics corresponding to credit investigation data of the plurality of sample users as input of the comprehensive credit assessment model to be trained, and determining predicted credit scores of the plurality of sample users through the comprehensive credit assessment model to be trained;
determining a comprehensive prediction error of the comprehensive credit evaluation model to be trained according to the prediction credit scores of the plurality of sample users and the sample credit scores;
according to a back propagation algorithm, performing back propagation on the comprehensive prediction error so as to update the model parameters of the comprehensive credit evaluation model to be trained;
and determining the comprehensive credit evaluation model to be trained after the model parameters are updated as the comprehensive credit evaluation model.
6. The method of claim 5, wherein determining the predictive credit scores for the plurality of sample users via the composite credit assessment model to be trained comprises:
evaluating first credit investigation characteristics respectively corresponding to credit investigation data of the plurality of sample users through an FM model to be trained and a DNN model to be trained in the comprehensive credit evaluation model to be trained to obtain first prediction credit scores and second prediction credit scores of the plurality of sample users;
and performing fusion processing on the first prediction credit scores and the second prediction credit scores of the plurality of sample users through an output layer in the comprehensive credit evaluation model to be trained to obtain the prediction credit scores of the plurality of sample users.
7. The method of claim 5, wherein the sample credit scores of the plurality of sample users comprise a plurality of credit scores in one-to-one correspondence with the plurality of types of credit investigation data;
determining a composite prediction error based on the prediction credit scores of the plurality of sample users and the sample credit scores, comprising:
determining a prediction error corresponding to each credit score according to the prediction credit scores of the plurality of sample users and each credit score in the plurality of credit scores to obtain a plurality of prediction errors;
determining weights for the plurality of prediction errors;
determining the composite prediction error based on the plurality of prediction errors and the weight of each prediction error.
8. An apparatus for determining credit data, the apparatus comprising:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring first credit investigation characteristics of credit investigation data of a target user, and the credit investigation data comprises various types of credit investigation data;
the evaluation module is used for taking the first credit investigation feature as an input of a comprehensive credit evaluation model, and evaluating the first credit investigation feature through an FM model and a DNN model in the comprehensive credit evaluation model respectively to obtain a first credit score and a second credit score of the target user;
and the processing module is used for carrying out fusion processing on the first credit score and the second credit score through an output layer of the comprehensive credit evaluation model to obtain a third credit score of the target user.
9. An apparatus for determining credit data, the apparatus comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to perform the steps of any of the methods of claims 1-7.
10. A computer-readable storage medium having instructions stored thereon, wherein the instructions, when executed by a processor, implement the steps of any of the methods of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911147899.9A CN110889759A (en) | 2019-11-21 | 2019-11-21 | Credit data determination method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911147899.9A CN110889759A (en) | 2019-11-21 | 2019-11-21 | Credit data determination method, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110889759A true CN110889759A (en) | 2020-03-17 |
Family
ID=69748218
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911147899.9A Pending CN110889759A (en) | 2019-11-21 | 2019-11-21 | Credit data determination method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110889759A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111582645A (en) * | 2020-04-09 | 2020-08-25 | 上海淇毓信息科技有限公司 | APP risk assessment method and device based on factorization machine and electronic equipment |
| CN112819610A (en) * | 2021-02-26 | 2021-05-18 | 重庆度小满优扬科技有限公司 | Credit evaluation method, credit evaluation model training method and device |
| CN113052693A (en) * | 2021-06-02 | 2021-06-29 | 北京轻松筹信息技术有限公司 | Data processing method and device, electronic equipment and computer readable storage medium |
| CN113159915A (en) * | 2021-02-25 | 2021-07-23 | 华夏方圆信用评估有限公司 | Intelligent financial credit dynamic evaluation method and system based on big data |
| CN113657993A (en) * | 2021-08-19 | 2021-11-16 | 中国平安财产保险股份有限公司 | Credit risk identification method, device, equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106127363A (en) * | 2016-06-12 | 2016-11-16 | 腾讯科技（深圳）有限公司 | A kind of user credit appraisal procedure and device |
| CN110175857A (en) * | 2019-03-25 | 2019-08-27 | 阿里巴巴集团控股有限公司 | It is preferred that business determines method and device |
| CN110264093A (en) * | 2019-06-21 | 2019-09-20 | 深圳前海微众银行股份有限公司 | Method for building up, device, equipment and the readable storage medium storing program for executing of Credit Model |
| CN110399979A (en) * | 2019-06-17 | 2019-11-01 | 深圳大学 | Click rate pre-estimation system and method based on field programmable gate array |
-
2019
- 2019-11-21 CN CN201911147899.9A patent/CN110889759A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106127363A (en) * | 2016-06-12 | 2016-11-16 | 腾讯科技（深圳）有限公司 | A kind of user credit appraisal procedure and device |
| CN110175857A (en) * | 2019-03-25 | 2019-08-27 | 阿里巴巴集团控股有限公司 | It is preferred that business determines method and device |
| CN110399979A (en) * | 2019-06-17 | 2019-11-01 | 深圳大学 | Click rate pre-estimation system and method based on field programmable gate array |
| CN110264093A (en) * | 2019-06-21 | 2019-09-20 | 深圳前海微众银行股份有限公司 | Method for building up, device, equipment and the readable storage medium storing program for executing of Credit Model |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111582645A (en) * | 2020-04-09 | 2020-08-25 | 上海淇毓信息科技有限公司 | APP risk assessment method and device based on factorization machine and electronic equipment |
| CN111582645B (en) * | 2020-04-09 | 2024-02-27 | 上海淇毓信息科技有限公司 | APP risk assessment method and device based on factoring machine and electronic equipment |
| CN113159915A (en) * | 2021-02-25 | 2021-07-23 | 华夏方圆信用评估有限公司 | Intelligent financial credit dynamic evaluation method and system based on big data |
| CN113159915B (en) * | 2021-02-25 | 2023-12-01 | 华夏方圆信用评估有限公司 | Intelligent financial credit dynamic assessment method and system based on big data |
| CN112819610A (en) * | 2021-02-26 | 2021-05-18 | 重庆度小满优扬科技有限公司 | Credit evaluation method, credit evaluation model training method and device |
| CN112819610B (en) * | 2021-02-26 | 2023-05-30 | 重庆度小满优扬科技有限公司 | Credit evaluation method, credit evaluation model training method and equipment |
| CN113052693A (en) * | 2021-06-02 | 2021-06-29 | 北京轻松筹信息技术有限公司 | Data processing method and device, electronic equipment and computer readable storage medium |
| CN113657993A (en) * | 2021-08-19 | 2021-11-16 | 中国平安财产保险股份有限公司 | Credit risk identification method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110889759A (en) | Credit data determination method, device and storage medium | |
| CN113408743B (en) | Method and device for generating federal model, electronic equipment and storage medium | |
| CN111695415A (en) | Construction method and identification method of image identification model and related equipment | |
| US20190354688A1 (en) | System and method for machine learning architecture with adversarial attack defence | |
| CN111177473B (en) | Personnel relationship analysis method, device and readable storage medium | |
| CN109766454A (en) | A kind of investor's classification method, device, equipment and medium | |
| CN114331829A (en) | Countermeasure sample generation method, device, equipment and readable storage medium | |
| US20210383205A1 (en) | Taxonomy Construction via Graph-Based Cross-domain Knowledge Transfer | |
| CN113011895B (en) | Associated account sample screening method, device and equipment and computer storage medium | |
| CN112785005B (en) | Multi-objective task assistant decision-making method and device, computer equipment and medium | |
| CN116010684A (en) | Article recommendation method, device and storage medium | |
| CN107368499B (en) | Client label modeling and recommending method and device | |
| CN112818995B (en) | Image classification method, device, electronic equipment and storage medium | |
| CN113822315A (en) | Attribute graph processing method and device, electronic equipment and readable storage medium | |
| CN113628059A (en) | Associated user identification method and device based on multilayer graph attention network | |
| CN112765481B (en) | Data processing method, device, computer and readable storage medium | |
| CN113449188A (en) | Application recommendation method and device, electronic equipment and readable storage medium | |
| CN108038496B (en) | Marriage and love object matching data processing method and device based on big data and deep learning, computer equipment and storage medium | |
| CN116362894A (en) | Multi-objective learning method, multi-objective learning device, electronic equipment and computer readable storage medium | |
| CN115600818A (en) | Multi-dimensional scoring method and device, electronic equipment and storage medium | |
| CN116258923A (en) | Image recognition model training method, device, computer equipment and storage medium | |
| CN115758271A (en) | Data processing method, data processing device, computer equipment and storage medium | |
| CN114820085B (en) | User screening method, related device and storage medium | |
| Nepomuceno | Multiobjective learning in the random neural network | |
| CN116881575B (en) | Content pushing method, device, computer equipment and storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200317 |
|
| RJ01 | Rejection of invention patent application after publication |