CN113537666B - Evaluation model training method, evaluation and business auditing method, device and equipment - Google Patents
Evaluation model training method, evaluation and business auditing method, device and equipment Download PDFInfo
- Publication number
- CN113537666B CN113537666B CN202010299090.4A CN202010299090A CN113537666B CN 113537666 B CN113537666 B CN 113537666B CN 202010299090 A CN202010299090 A CN 202010299090A CN 113537666 B CN113537666 B CN 113537666B
- Authority
- CN
- China
- Prior art keywords
- evaluation
- model
- sub
- data
- user
- 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.)
- Active
Links
- 238000013210 evaluation model Methods 0.000 title claims abstract description 147
- 238000011156 evaluation Methods 0.000 title claims abstract description 132
- 238000000034 method Methods 0.000 title claims abstract description 74
- 238000012549 training Methods 0.000 title claims abstract description 51
- 238000012502 risk assessment Methods 0.000 claims abstract description 7
- 238000012550 audit Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 abstract description 10
- 230000004927 fusion Effects 0.000 description 18
- 230000008569 process Effects 0.000 description 13
- 238000010586 diagram Methods 0.000 description 9
- 230000000694 effects Effects 0.000 description 7
- 230000005540 biological transmission Effects 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 230000002093 peripheral effect Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Biology (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Development Economics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Educational Administration (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention discloses an evaluation model training method, an evaluation and business auditing device and equipment, relates to the technical field of data processing, and aims to solve the problem of low accuracy of risk evaluation. The method comprises the following steps: acquiring sample data; inputting the sample data into a pre-evaluation model to obtain a pre-evaluation score of the sample data; extracting intermediate sample data of the sample data based on the pre-evaluation score; model training is carried out based on the intermediate sample data, and a risk evaluation model of the intermediate sample data is obtained; the pre-evaluation score of the intermediate sample data is larger than a first preset value and smaller than a second preset value, and the first preset value is smaller than the second preset value; the first preset value and the second preset value are both greater than 0. The embodiment of the invention can improve the accuracy of risk assessment.
Description
Technical Field
The present invention relates to the field of data processing technologies, and in particular, to an evaluation model training method, an evaluation and service auditing method, an apparatus, a device, and a storage medium.
Background
The wind control capability is the core competitiveness in the financial field, and the promotion of the wind control capability is not separated from the technical support of big data. With the popularization of the internet, particularly, the mobile internet, various information of everyone in life is digitally recorded, such as illegal information in public security, credit information in credit, call record information in operator service, and the like. These data each represent a customer from different dimensions, and the overall risk of the customer requires integration of the information from each dimension for comprehensive evaluation.
Along with the improvement of data dimension, for facilitating approval use, the use of related data features is generally considered to train a model, and comprehensive risk assessment is carried out on clients through the model, so that judgment is made on the final approval passing rejection.
However, in the prior art, when performing risk assessment, the best guest group and the worst guest group can be distinguished through the model, but the middle guest group cannot be distinguished accurately, so that the accuracy of risk assessment is not high.
Disclosure of Invention
The embodiment of the invention provides an evaluation model training method, an evaluation and business auditing device, equipment and a storage medium, which are used for solving the problem of low accuracy of risk evaluation.
In a first aspect, an embodiment of the present invention provides an evaluation model training method, including:
Acquiring sample data;
Inputting the sample data into a pre-evaluation model to obtain a pre-evaluation score of the sample data;
Extracting intermediate sample data of the sample data based on the pre-evaluation score, wherein the pre-evaluation score of the intermediate sample data is larger than a first preset value and smaller than a second preset value, the first preset value is smaller than the second preset value, and the first preset value and the second preset value are both larger than 0;
and inputting the intermediate sample data into a neural network for model training to obtain a risk evaluation model of the intermediate sample data.
In a second aspect, an embodiment of the present invention further provides an evaluation method, including:
acquiring user information of a user to be evaluated;
According to the user information, determining guest group information corresponding to the user to be evaluated;
Based on a pre-evaluation model corresponding to the guest group information, obtaining a pre-evaluation score of the user to be evaluated;
Inputting the user information into a risk evaluation model under the condition that the pre-evaluation score is larger than a first preset value and smaller than a second preset value to obtain the risk evaluation score of the user to be evaluated;
wherein the first preset value and the second preset value are both greater than 0.
In a third aspect, an embodiment of the present invention further provides a service auditing method, including:
Receiving a service request of a user;
acquiring user information according to the service request of the user;
Evaluating the user by using the evaluating method according to any one of claim 7 or claim 8 based on the user information to obtain a risk evaluating score of the user;
And obtaining a business auditing result of the user according to the risk evaluating score.
In a fourth aspect, an embodiment of the present invention further provides a service auditing apparatus, including:
The first receiving module is used for receiving the service request of the user;
The first acquisition module is used for acquiring user information according to the service request of the user;
the first evaluation module is used for evaluating the user by using the evaluation method according to any one of the third aspect based on the user information to obtain a risk evaluation score of the user;
and the second acquisition module is used for acquiring a business auditing result of the user according to the risk evaluation score.
In a fifth aspect, an embodiment of the present invention further provides an electronic device, including: a memory, a processor and a program stored on the memory and executable on the processor, the processor implementing steps comprising the method for training an evaluation model as described in the first aspect when the program is executed; or to implement the steps comprising the evaluation method as described in the second aspect; or to implement the steps of the evaluation method comprising the third aspect.
In a sixth aspect, embodiments of the present invention further provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps comprising the method for training an evaluation model according to the first aspect; or to implement the steps comprising the evaluation method as described in the second aspect; or to implement the steps of the evaluation method comprising the third aspect.
In the embodiment of the invention, the user information of the user to be evaluated is used for obtaining the pre-evaluation score of the user to be evaluated through the pre-evaluation model of the corresponding guest group information. And if the pre-evaluation score is larger than a first preset value and smaller than a second preset value, obtaining the risk evaluation score of the user to be evaluated based on the user information and the risk evaluation model, so as to determine whether the user to be evaluated passes the risk evaluation. Under the condition that the pre-evaluation score is larger than a first preset value and smaller than a second preset value, the user to be evaluated is evaluated again, so that different guest groups can be distinguished accurately, and the accuracy of an evaluation result is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments of the present invention will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort to a person of ordinary skill in the art.
FIG. 1 is one of the flowcharts of the method for training an evaluation model provided by the embodiment of the invention;
FIG. 2 is one of the flowcharts of the evaluation method provided by the embodiment of the present invention;
FIG. 3 is a second flowchart of an evaluation method according to an embodiment of the present invention;
FIG. 4 is a third flowchart of an evaluation method according to an embodiment of the present invention;
FIG. 5 is a flowchart of a business audit method provided by an embodiment of the present invention;
FIG. 6 is a block diagram of an evaluation model training device provided by an embodiment of the present invention;
FIG. 7 is a block diagram of an evaluation device according to an embodiment of the present invention;
FIG. 8 is a block diagram of a business audit device according to an embodiment of the present invention;
FIG. 9 is one of the block diagrams of the electronic device provided by the embodiment of the invention;
FIG. 10 is a second block diagram of an electronic device according to an embodiment of the present invention;
fig. 11 is a third block diagram of the electronic device according to the embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, fig. 1 is a flowchart of an evaluation model training method provided by an embodiment of the present invention, as shown in fig. 1, including the following steps:
and 101, acquiring sample data.
In the embodiment of the invention, in order to improve the accuracy of the evaluation result, the sample data comprises sub-sample data of at least one data source. That is, the sample data in embodiments of the present invention may come from different data sources. The data of the different data sources is referred to herein as sub-sample data.
For example, the data source may be infraction information in public security, credit information in credit, call log information in operator service, etc.
And 102, inputting the sample data into a pre-evaluation model to obtain a pre-evaluation score of the sample data.
The pre-evaluation model can be realized by using a risk evaluation model in the prior art, and can also be trained by the following method according to the embodiment of the invention. The role of the pre-evaluation model is to distinguish between different users based on preset goals. For example, the preset target may be a payment overdue rate, and then the payment overdue rates of different users may be obtained through the pre-evaluation model, so as to divide different guest groups.
In the embodiment of the invention, the pre-evaluation model can be obtained as follows.
(1) For each sub-sample data, each sub-sample data is divided into at least two guest group data, respectively.
In the embodiment of the invention, the sub-sample data of each data source can be divided into at least two guest group data according to different classification standards.
For example, for sub-sample data of a data source, the sub-sample data may be divided into a loan-with-house group and a loan-without-house group for the feature of "loan-with-house" in credit. For another example, the sub-sample data may be divided into credit record with credit and credit record without credit record for credit sign.
(2) And obtaining a sub pre-evaluation model of each group of guest data.
In this step, a sub-model of each guest group data is determined.
Taking first guest group data in each guest group data as an example, first, a sub-model corresponding to the first guest group data is determined for each first guest group data. The first guest data is any one of all guest data of all data sources. And then, based on the sub-model corresponding to the first customer group data and a first preset algorithm, obtaining a sub-pre-evaluation model of the first customer group data. The first preset algorithm may be, for example, a linear model, a tree model, or the like.
For each group of data for each data source, there are one or more sub-models in embodiments of the invention, depending on the model objectives. The model object, i.e. the role of the model, is, for example, whether the model is used for identifying medium-short period overdue clients or long period overdue clients, etc.
Thus, in the embodiment of the present invention, taking the first customer group data as an example, the target data is selected from the first customer group data; the target data corresponds to a model target, and then a sub-model of the first customer group data is obtained based on the target data and a second preset algorithm. The second preset algorithm may be, for example, a linear model, a tree model, or the like.
For example, if the model target is for evaluating the expiration rate of a short period, here, data of a billing period between MOB (moth on book, at billing month) 2-MOB4 may be selected from the first class group data as target data.
In practical applications, in order to identify both short-period overdue client features and long-period overdue client features in an approval, in an embodiment of the present invention, multiple sub-models for different data sources may be trained using overdue targets (e.g., FPD20, etc.) that represent short periods and overdue targets (e.g., dpd90@mob12, etc.) that represent long periods, respectively. In practical applications, the characteristics of the guest groups in different time periods may have certain differences. Thus, in model training, samples with an account age between MOB2-MOB4 are used for mid-short period model training; for long period model training, samples with an account age between MOB4-MOB12 were used. Therefore, in the sub-sample data of each data source, the guest group of the nearest month meeting the condition is extracted from the samples meeting the requirement of account age for model training, so that the guest group difference caused by time lag is reduced to the greatest extent.
(3) And fusing the sub pre-evaluation models of the guest group data to obtain the pre-evaluation model.
Here, the meaning of "fusion" refers to comprehensively considering the output of each sub-pre-evaluation model to obtain a final pre-evaluation model. And obtaining the pre-evaluation model based on the output of the sub-pre-evaluation models of all the guest groups and a third preset algorithm. The third preset algorithm may be, for example, a linear model, a tree model, or the like.
In this step, different guest group data of different data sources are input into the pre-evaluation model, so as to obtain pre-evaluation scores of different guest groups of different sample data. And comparing the pre-evaluation score with a preset value to obtain pre-evaluation results of different guest groups.
Step 103, extracting middle sample data of the sample data based on the pre-evaluation score.
Based on the comparison result, the sample data can be divided into three types: pre-evaluating sample data meeting requirements, middle sample data, and pre-evaluating sample data not meeting requirements. The pre-evaluation score of the intermediate sample data is larger than a first preset value and smaller than a second preset value, and the first preset value is smaller than the second preset value; the sample data with the pre-evaluation score being greater than or equal to the second preset value is sample data with the pre-evaluation meeting the requirements, and the sample data with the pre-evaluation score being less than or equal to the first preset value is sample data with the pre-evaluation not meeting the requirements; the first preset value and the second preset value are both greater than 0.
The three sample data may include data for different guest groups in different data sources. That is, a certain class of sample data may come from different clusters of different data sources. In the embodiment of the invention, the intermediate sample data is extracted for subsequent further evaluation.
And 104, inputting the intermediate sample data into a neural network for model training to obtain a risk evaluation model of the intermediate sample data.
In this step, target group data corresponding to the intermediate sample data is determined from at least two group data of each data source. And then, a sub-pre-evaluation model of the target guest group data is acquired. And finally, fusing the sub pre-evaluation models of the target guest group data to obtain a risk evaluation model of the intermediate sample data.
Since the intermediate sample data may come from different data sources, in the embodiment of the present invention, the data source corresponding to the intermediate sample data may be determined according to the correspondence between the intermediate sample data and the data source. Different data sources are divided into different guest groups, and then the guest groups corresponding to the intermediate sample data can be determined according to the relationship between the data sources and the guest groups. According to the above description, different guest groups of different data sources have corresponding sub-pre-evaluation models, so the sub-pre-evaluation models of the target guest group data corresponding to the intermediate sample data can be determined according to the corresponding relation. And then, fusing the sub pre-evaluation models of the target guest group data to obtain a risk evaluation model of the intermediate sample data.
In the embodiment of the invention, the meaning of fusion is that a plurality of sub-pre-evaluation models are connected together through a certain algorithm. For example, the outputs of the plurality of sub-predictive models may be calculated by means of linear weighting, so as to obtain a risk evaluation model.
In the embodiment of the invention, the user information of the user to be evaluated is used for obtaining the pre-evaluation score of the user to be evaluated through the pre-evaluation model of the corresponding guest group information. And if the pre-evaluation score is larger than a first preset value and smaller than a second preset value, obtaining the risk evaluation score of the user to be evaluated based on the user information and the risk evaluation model, so as to determine whether the user to be evaluated passes the risk evaluation. Under the condition that the pre-evaluation score is larger than a first preset value and smaller than a second preset value, the user to be evaluated is evaluated again, so that different guest groups can be distinguished accurately, and the accuracy of an evaluation result is improved.
Referring to fig. 2, fig. 2 is a flowchart of an evaluation method provided by an embodiment of the present invention, as shown in fig. 2, including the following steps:
Step 201, obtaining user information of a user to be evaluated.
The user information of the user to be evaluated can include, for example, age, sex, business processing record, loan record, credit information and the like of the user.
Step 202, determining guest group information corresponding to the user to be evaluated according to the user information.
For the sub-sample data of a certain data source, the characteristic of whether a house is lended or not in credit investigation can be divided into a house-lending group and a house-free lending group. As another example, for "credit record with or without credit" in credit, credit record with or without credit record guest groups may be divided. Therefore, the customer group information corresponding to the user to be evaluated can be determined according to the dividing standard.
And 203, obtaining the pre-evaluation score of the user to be evaluated based on the pre-evaluation model corresponding to the guest group information.
In the embodiment of the invention, the pre-evaluation model corresponding to each guest group data in different data sources can be obtained according to the method. Thus, a corresponding pre-evaluation model may be determined based on the determined guest group information. The processing efficiency can be improved by selecting the corresponding pre-evaluation model based on the guest group information of the user to be evaluated.
In this step, a sub-model score is obtained based on the sub-model corresponding to the user information and the guest group information. And then, based on the sub-model score and a sub-pre-evaluation model corresponding to the guest group information, obtaining a sub-pre-evaluation model score corresponding to the user information. And finally, obtaining the pre-evaluation score of the user to be evaluated based on the sub-pre-evaluation model score and the pre-evaluation model.
For example, after the sub-model score is obtained, the sub-model score is input into a corresponding sub-pre-evaluation model to obtain the sub-pre-evaluation model score. Then, each sub-pre-evaluation model score is input into the pre-evaluation model to obtain a pre-evaluation score.
Based on different pre-evaluation scores, the category of the user to be evaluated can be judged.
In the case where the pre-evaluation score is greater than the first preset value and less than the second preset value (the first preset value is less than the second preset value, both of which are greater than 0), the user to be evaluated can be considered as a client who needs to evaluate it again. If the value is larger than the second preset value, the approval can be directly considered to pass. If the value is smaller than the first preset value, the approval can be directly considered to be refused.
And 204, inputting the user information into a risk evaluation model to obtain the risk evaluation score of the user to be evaluated and obtain the risk evaluation score of the user to be evaluated under the condition that the pre-evaluation score is larger than a first preset value and smaller than a second preset value.
The risk evaluation model is obtained by the evaluation model training method.
In the case that the pre-evaluation score is greater than the first preset value and less than the second preset value, the user to be evaluated may be considered as a client who needs to evaluate it again. At this time, the risk evaluation model obtained by training by the method can be utilized to obtain the risk evaluation score of the user to be evaluated.
And under the condition that the risk evaluation score is greater than or equal to a preset threshold value, determining that the user to be evaluated passes through risk evaluation. If the risk is smaller than the preset threshold, confirming that the risk evaluation is not passed. Wherein the preset threshold is an empirically set value.
In the embodiment of the invention, the user information of the user to be evaluated is used for obtaining the pre-evaluation score of the user to be evaluated through the pre-evaluation model of the corresponding guest group information. And if the pre-evaluation score is larger than a first preset value and smaller than a second preset value, obtaining the risk evaluation score of the user to be evaluated based on the user information and the risk evaluation model, so as to determine whether the user to be evaluated passes the risk evaluation. Under the condition that the pre-evaluation score is larger than a first preset value and smaller than a second preset value, the user to be evaluated is evaluated again, so that different guest groups can be distinguished accurately, and the accuracy of an evaluation result is improved.
Referring to fig. 3, fig. 3 is a flowchart of an evaluation method provided by an embodiment of the present invention, as shown in fig. 3, including the following steps:
step 301, acquiring data of a plurality of data sources.
For the stock clients, the data of the data source A-data source N is obtained by querying the relevant data sources. Wherein the data source may be credit data, illegal records, consumption records, etc.
Step 302, dividing the data of each data source into data of different guest groups.
According to analysis, the same data source has different characteristics of the guest group used for identifying the short period overdue and the long period overdue of the client, and the guest group characteristics of different time periods have certain changes along with the time. Different guest groups have significant differences in data source characteristics. Therefore, the guest groups can be subdivided according to specific characteristics, and a targeted model is built for the specific guest groups.
For example, the guest groups are classified into a loan guest group and a loan guest group according to the feature of "whether there is a loan" in the credit investigation. Wherein, the liabilities of the groups of house lenders are relatively high, and the credit repayment records are relatively good. For another example, the guest groups are divided into credit-record guest groups and credit-record-free guest groups according to "credit-record-not-present" in credit assessment. Wherein, the credit record guest group is relatively less in risk because of the approval by the bank, and the credit record guest group is not available because of less information available in the approval, and the risk assessment of the partial persons is relatively weak.
Step 303, training a sub-model for different guest group data of different data sources.
In order to identify both short-period overdue client features and long-period overdue client features in an approval, multiple sub-models are trained for different clusters of different data sources using overdue targets representing short periods (e.g., FPD20, etc.) and overdue targets representing long periods (e.g., dpd90@mob12, etc.), respectively. At the same time, since fraud samples have been identified in the pre-credit approval, fraud-related sub-models of the data sources may also be trained herein.
Specifically, due to economic policy adjustment, the situation that the credit customers sink as a whole occurs, and the characteristics of the customers in different time periods may have certain differences. Thus, in model training, for short period model training, use is made of samples of account age between MOB2-MOB 4; for long period model training, account age was taken between MOB4-MOB 12. Based on the principle, the passenger group of the latest month meeting the conditions is extracted from the passenger group data to carry out model training, so that the passenger group difference caused by time lag is reduced to the greatest extent.
And training sub-models of the data sources aiming at different guest groups and different targets by using algorithms such as a tree model, a linear model and the like respectively by using the subdivided guest group data and different model targets and the guest group data screened according to the time period.
The same guest group may have multiple sub-models. For example, the submodel of data source a may include submodels A1, A2 … …; the submodels of the data source N may include submodels N1, N2 … …, and the like.
Step 304, training a sub-pre-evaluation model for different guest group data of different data sources.
Wherein the sub-pre-evaluation model may also be referred to as a sub-fusion model.
In step 303, the output of the sub-model, i.e., the sub-model score, may be obtained. Thus, in this step, the sub-fusion models of the sub-models for different guest groups and different targets can be trained by dividing the sub-models into input features for different guest groups.
In the training of the sub-fusion model, both long and short period sub-models can be used as input features. Different guest groups can respectively screen corresponding guest group sample data, and different algorithms (such as a linear model, a tree model and the like) are used for training the sub-fusion model.
And 305, obtaining a fusion model by using the sub pre-evaluation model.
In step 304, the output of the sub-fusion model, i.e., the sub-fusion model score, may be obtained. And in different guest groups, taking the output of the guest group corresponding sub fusion model as an input characteristic, and using a tree model or a linear model to obtain a fusion model.
And 306, training a risk evaluation model aiming at the middle score guest group.
In analyzing the guest group using the fusion model described above, it was found that good samples (model score higher (20% by weight), e.g., greater than a first threshold), bad samples (model score lower (20% by weight), e.g., less than a second threshold), and intermediate samples (model score between the second threshold and the first threshold, the first threshold being greater than the second threshold) can be distinguished by the model of the fusion model.
In order to increase the degree of discrimination of the intermediate samples in the model division, the intermediate samples are screened for each guest group on the basis of the fusion model, and the sub-model division of the intermediate samples is independently used as a characteristic for model training, so that a model of the intermediate samples, namely a risk evaluation model, is generated.
According to analysis, the correlation between the model of the intermediate sample and the total guest group fusion model is 15%, and the guest group of the intermediate sample is refined, so that good samples and bad samples in the intermediate sample can be accurately distinguished.
The risk evaluation model of the intermediate sample is similar to the training process of the fusion model. The sub-fusion model may be trained using the respective sub-model sub-models of the guest group to which the intermediate sample corresponds, and then the intermediate fusion model of the corresponding guest group may be trained using the sub-fusion model sub-models.
According to the above process, when a new customer applies for loan approval, firstly, each data source information is queried, and the new customer is classified into the corresponding customer group according to the obtained characteristics. And then, analyzing the client by using a model corresponding to the client group, thereby simplifying the approval process and improving the approval efficiency.
Specifically, after the data characteristic information of the new client is transmitted to the trained data source sub-model, each sub-model is obtained. The sub-model components are transmitted to each sub-fusion model as features, and corresponding sub-fusion model components are obtained. And each sub-fusion model score is used as the input of the fusion model to obtain the fusion model score.
If the model is greater than or equal to A, the client is in a good group and directly passes approval; if the model score is less than or equal to B, indicating that the client is in a bad client group, and directly refusing approval; if the model score is larger than B and smaller than A, the intermediate client model is used for prediction, and passing or rejecting is judged according to a specific threshold and the passing rate set in approval. Wherein A, B is greater than 0, and A is greater than B.
Taking the loan application process of the new customer as an example, referring to fig. 4, the relevant data information of the new customer, such as credit information data, bank card data, multi-platform loan data, blacklist data and other data sources 1-4, is queried first.
The characteristics of the guest groups are analyzed to categorize new guests into specific guest groups. If the new client has credit transaction information, if so, dividing the new client into the client group with credit transaction. The corresponding models which are used subsequently are models which are obtained by training on credit transaction samples, so that the model effect can be improved, and meanwhile, the approval process efficiency is improved.
After classifying the new clients into the specific client group, respectively calculating the models which are trained by each data source for the client group, and obtaining corresponding sub-model components.
For example, in the trained credit sub-models A1-An, credit data is input into a credit sub-model corresponding to guest group information of a new customer, and sub-model scores are obtained. And in the trained bank clip models B1-Bn, the bank card data are input into the bank clip model corresponding to the customer group information of the new customer, so as to obtain the sub-model score. And so on.
And using the corresponding small model score as a characteristic, inputting the characteristic into a sub-fusion model of a guest group corresponding to the new customer to obtain a corresponding sub-fusion model score s1/s2 … and the like.
Aiming at the sub-fusion model parts, a linear weighting mode is used to obtain fusion model parts. For example, for the obtained s1 to sN sub-fusion model components, the sub-fusion model components may be weighted linearly by geometric average, or the sub-fusion model components may be multiplied by the corresponding weights, and then the result of the multiplication may be subjected to summation operation or the like.
And dividing the new client into client groups at the model layer according to the fusion model division. For example, if the model score is greater than or equal to A, it indicates that the client is in a good group, and the client is directly approved; if the model score is less than or equal to B, indicating that the client is in a bad client group, and directly refusing approval; if the model score is greater than B and less than A, the trained intermediate guest group model (risk evaluation model) is used, the model score of the guest group is calculated again, and the passing or rejecting is judged according to a specific threshold and the passing rate set in the approval. Wherein A, B is greater than 0, and A is greater than B.
If the model score is larger than a preset threshold value, the approval passes; otherwise, the approval is refused. It should be noted that the threshold may be changed according to the change of the approval passing rate, for example, the passing rate may be improved, and the threshold may be reduced.
As can be seen from the above description, in the embodiment of the present invention, since the guest groups are refined, and different guest groups correspond to different approval flows, by using the scheme of the embodiment of the present invention, flow differentiation of different guest groups can be achieved, and overall approval flow efficiency of the client is improved. Meanwhile, aiming at guest group refinement, different guest groups use sample training models corresponding to guest groups, the model effect is obviously superior to that of a model trained by all samples of coarse guang style, and the model method has greater advantages in screening clients.
In the embodiment of the invention, the risk evaluation model is independently trained for the customer groups with the fused model scores in the middle section, so that the distinction between good customers and bad customers in the section is refined. When a new customer falls into the middle interval, the model is used for predicting again, so that the accuracy of evaluation is improved.
In the process of utilizing the embodiment of the invention, if the single data source is abnormal in the test process, the model effect attenuation is obviously lower than that of the traditional approval process to construct the model by utilizing the scheme of the embodiment of the invention. The method is characterized in that the data source suddenly appears abnormal, the fusion model is trained by using the bottom layer features, the number of the occurrence of the missing features is large, and the whole model is greatly deviated. The sub-model components are used as variables to train the fusion model, the feature number is reduced, the sub-model component value range is between 0 and 1, and 0.5 is used as a default substitution value, so that the deviation of the fusion model components is reduced to a large extent.
Referring to fig. 5, fig. 5 is a flowchart of an evaluation method provided by an embodiment of the present invention, as shown in fig. 5, including the following steps:
step 501, a service request of a user is received.
The user's business request may represent different meanings depending on the application scenario. Such as when the embodiments of the present invention are applied to a pre-credit approval scenario, the user's business request may represent a request from the user to apply for a loan.
Step 502, obtaining user information according to the service request of the user.
The user information may include, for example, a user name, an age, repayment information, historical consumption information, and the like. Also, different user information may be available in different scenarios.
And step 503, evaluating the user by using an evaluating method based on the user information to obtain a risk evaluating score of the user.
The evaluation method is the evaluation method.
And 504, obtaining a business auditing result of the user according to the risk evaluation score.
According to the description of the embodiment of the method, if the obtained risk evaluation score is greater than or equal to A, the user is directly approved; if the risk evaluation score is smaller than or equal to B, directly refusing approval; if the risk evaluation score is greater than B and less than A, the trained middle guest group model (risk evaluation model) is used to calculate the risk evaluation score of the user again, and pass or reject is judged according to a specific threshold and the passing rate set in approval. If the risk evaluation score is greater than a preset threshold value, passing the approval; otherwise, the approval is refused. Wherein A, B is greater than 0, and A is greater than B.
It should be noted that the solution according to the embodiment of the present invention may be applied in various scenarios, such as pre-credit approval, marketing, anti-fraud, credit, etc.
By the scheme provided by the embodiment of the invention, the user is evaluated through the obtained risk evaluation score, so that different guest groups can be accurately distinguished, and the accuracy of an evaluation result is improved.
The embodiment of the invention also provides an evaluation model training device. Referring to fig. 6, fig. 6 is a structural diagram of an evaluation model training device provided by an embodiment of the present invention. Because the principle of solving the problem of the evaluation model training device is similar to that of the evaluation model training method in the embodiment of the invention, the implementation of the evaluation model training device can be referred to the implementation of the method, and the repetition is omitted.
As shown in fig. 6, the evaluation model training apparatus 600 includes:
A first obtaining module 601, configured to obtain sample data; a second obtaining module 602, configured to input the sample data to the pre-evaluation model, and obtain a pre-evaluation score of the sample data; a first extraction module 603 for extracting intermediate sample data of the sample data based on the pre-evaluation score; the pre-evaluation score of the intermediate sample data is larger than a first preset value and smaller than a second preset value, and the first preset value is smaller than the second preset value; the first preset value and the second preset value are both larger than 0; the training module 604 is configured to input the intermediate sample data to a neural network for model training, so as to obtain a risk evaluation model of the intermediate sample data;
the sample data with the pre-evaluation score being greater than or equal to the second preset value is sample data which is pre-evaluated to meet the requirements, and the sample data with the pre-evaluation score being less than or equal to the first preset value is sample data which is pre-evaluated to not meet the requirements; the first preset value and the second preset value are both greater than 0.
Optionally, the sample data comprises sub-sample data of at least one data source; the apparatus may further include: and the third acquisition module is used for acquiring the pre-evaluation model. Wherein the third acquisition module may include:
The dividing sub-module is used for dividing each sub-sample data into at least two guest group data respectively; the first acquisition sub-module is used for acquiring a sub-pre-evaluation model of each guest group data; and the second acquisition sub-module is used for fusing the sub-pre-evaluation models of the guest group data to obtain the pre-evaluation models.
Optionally, the first obtaining sub-module may include:
A first determining unit, configured to determine, for each first customer group data, a sub-model corresponding to the first customer group data; the first guest data is any one of all guest data of all data sources; the first acquisition unit is used for acquiring a sub pre-evaluation model of the first customer group data based on the sub model corresponding to the first customer group data and a first preset algorithm.
Optionally, the first determining unit includes:
A first selection subunit configured to select target data from the first guest group data; wherein the target data corresponds to a model target; the first acquisition subunit is configured to obtain a sub-model of the first customer group data based on the target data and a second preset algorithm.
Optionally, the first obtaining subunit is configured to obtain the pre-evaluation model based on an output of the sub-pre-evaluation models of all the guest groups and a third preset algorithm.
Optionally, the training module 604 may include:
A determining submodule, configured to determine target group data corresponding to the intermediate sample data from at least two group data of each data source; the first acquisition sub-module is used for acquiring a sub-pre-evaluation model of the target guest group data; and the second acquisition sub-module is used for fusing the sub-pre-evaluation models of the target guest group data to obtain a risk evaluation model of the intermediate sample data.
The device provided by the embodiment of the present invention may execute the above method embodiment, and its implementation principle and technical effects are similar, and this embodiment will not be described herein.
The embodiment of the invention also provides an evaluation device. Referring to fig. 7, fig. 7 is a block diagram of an evaluation device provided in an embodiment of the present invention. Because the principle of the evaluating device for solving the problems is similar to that of the evaluating method in the embodiment of the invention, the implementation of the evaluating device can be referred to the implementation of the method, and the repetition is omitted.
As shown in fig. 7, the evaluation apparatus 700 includes:
A first obtaining module 701, configured to obtain user information of a user to be evaluated; a first determining module 702, configured to determine, according to the user information, guest group information corresponding to the user to be evaluated; a second obtaining module 703, configured to obtain a pre-evaluation score of the user to be evaluated based on a pre-evaluation model corresponding to the guest group information; a third obtaining module 704, configured to input the user information into a risk evaluation model to obtain a risk evaluation score of the user to be evaluated, where the pre-evaluation score is greater than a first preset value and less than a second preset value; wherein the first preset value and the second preset value are both greater than 0.
Wherein the pre-evaluation model and the risk evaluation model are obtained by using an evaluation model training method comprising the above.
Optionally, the second obtaining module 703 may include:
The first acquisition sub-module is used for obtaining a sub-model component based on the sub-model corresponding to the user information and the guest group information; the second acquisition sub-module is used for acquiring a sub-pre-evaluation model score corresponding to the user information based on the sub-model score and a sub-pre-evaluation model corresponding to the guest group information; and the third acquisition submodule is used for acquiring the pre-evaluation score of the user to be evaluated based on the sub-pre-evaluation model score and the pre-evaluation model.
Optionally, the apparatus may further include:
And the second determining module is used for determining that the user to be evaluated passes through risk evaluation under the condition that the risk evaluation score is greater than or equal to a preset threshold value.
The device provided by the embodiment of the present invention may execute the above method embodiment, and its implementation principle and technical effects are similar, and this embodiment will not be described herein.
The embodiment of the invention also provides a business auditing device. Referring to fig. 8, fig. 8 is a block diagram of a service auditing apparatus according to an embodiment of the present invention. Because the principle of the business auditing device for solving the problem is similar to that of the business auditing method in the embodiment of the invention, the implementation of the business auditing device can refer to the implementation of the method, and the repetition is omitted.
As shown in fig. 8, the business auditing apparatus 800 includes:
A first receiving module 801, configured to receive a service request of a user; a first obtaining module 802, configured to obtain user information according to a service request of the user; a first evaluation module 803, configured to evaluate the user by using an evaluation method as described above based on the user information, to obtain a risk evaluation score of the user; and a second obtaining module 804, configured to obtain a business audit result of the user according to the risk evaluation score.
The device provided by the embodiment of the present invention may execute the above method embodiment, and its implementation principle and technical effects are similar, and this embodiment will not be described herein.
As shown in fig. 9, an electronic device according to an embodiment of the present invention includes: processor 900, for reading the program in memory 920, performs the following procedures:
Acquiring sample data;
Inputting the sample data into a pre-evaluation model to obtain a pre-evaluation score of the sample data;
Extracting intermediate sample data of the sample data based on the pre-evaluation score, wherein the pre-evaluation score of the intermediate sample data is larger than a first preset value and smaller than a second preset value, the first preset value is smaller than the second preset value, and the first preset value and the second preset value are both larger than 0;
Model training is carried out based on the intermediate sample data, and a risk evaluation model of the intermediate sample data is obtained;
The sample data with the pre-evaluation score being greater than or equal to the second preset value is sample data which is pre-evaluated to meet the requirements, and the sample data with the pre-evaluation score being less than or equal to the first preset value is sample data which is pre-evaluated to not meet the requirements;
the first preset value and the second preset value are both greater than 0.
A transceiver 910 for receiving and transmitting data under the control of the processor 900.
Wherein in fig. 9, a bus architecture may comprise any number of interconnected buses and bridges, and in particular one or more processors represented by processor 900 and various circuits of memory represented by memory 920, linked together. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, power management circuits, etc., which are well known in the art and, therefore, will not be described further herein. The bus interface provides an interface. The transceiver 910 may be a number of elements, including a transmitter and a transceiver, providing a means for communicating with various other apparatus over a transmission medium. The processor 900 is responsible for managing the bus architecture and general processing, and the memory 920 may store data used by the processor 900 in performing operations.
The processor 900 is responsible for managing the bus architecture and general processing, and the memory 920 may store data used by the processor 900 in performing operations.
The sample data comprises sub-sample data of at least one data source; the processor 900 is further configured to read the program, and perform the following steps:
obtaining the pre-evaluation model comprises the following steps:
dividing each sub-sample data into at least two guest group data for each sub-sample data;
Obtaining a sub pre-evaluation model of each guest group data;
And fusing the sub pre-evaluation models of the guest group data to obtain the pre-evaluation model.
The processor 900 is further configured to read the program, and perform the following steps:
for each first customer group data, determining a sub-model corresponding to the first customer group data; the first guest data is any one of all guest data of all data sources;
And obtaining a sub pre-evaluation model of the first customer group data based on the sub model corresponding to the first customer group data and a first preset algorithm.
The processor 900 is further configured to read the program, and perform the following steps:
selecting target data from the first customer base data; wherein the target data corresponds to a model target;
and obtaining a sub-model of the first customer group data based on the target data and a second preset algorithm.
The processor 900 is further configured to read the program, and perform the following steps:
And obtaining the pre-evaluation model based on the output of the sub-pre-evaluation models of all the guest groups and a third preset algorithm.
The processor 900 is further configured to read the program, and perform the following steps:
Determining target group data corresponding to the intermediate sample data from at least two group data of each data source;
obtaining a sub-pre-evaluation model of the target guest group data;
and fusing the sub pre-evaluation models of the target guest group data to obtain a risk evaluation model of the intermediate sample data.
As shown in fig. 10, an electronic device according to an embodiment of the present invention includes: processor 1000, for reading the program in memory 1020, performs the following processes:
acquiring user information of a user to be evaluated;
According to the user information, determining guest group information corresponding to the user to be evaluated;
Based on a pre-evaluation model corresponding to the guest group information, obtaining a pre-evaluation score of the user to be evaluated;
Inputting the user information into a risk evaluation model under the condition that the pre-evaluation score is larger than a first preset value and smaller than a second preset value to obtain the risk evaluation score of the user to be evaluated;
wherein the first preset value and the second preset value are both greater than 0.
A transceiver 1010 for receiving and transmitting data under the control of the processor 1000.
Wherein in fig. 10, a bus architecture may comprise any number of interconnected buses and bridges, and in particular one or more processors represented by the processor 1000 and various circuits of the memory, represented by the memory 1020, are chained together. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, power management circuits, etc., which are well known in the art and, therefore, will not be described further herein. The bus interface provides an interface. The transceiver 1010 may be a number of elements, including a transmitter and a transceiver, providing a means for communicating with various other apparatus over a transmission medium. The processor 1000 is responsible for managing the bus architecture and general processing, and the memory 1020 may store data used by the processor 1000 in performing operations.
The processor 1000 is responsible for managing the bus architecture and general processing, and the memory 1020 may store data used by the processor 1000 in performing operations.
The processor 1000 is further configured to read the program and perform the following steps:
obtaining a sub-model score based on the sub-model corresponding to the user information and the guest group information;
Based on the sub-model score and a sub-pre-evaluation model corresponding to the guest group information, obtaining a sub-pre-evaluation model score corresponding to the user information;
And obtaining the pre-evaluation score of the user to be evaluated based on the sub-pre-evaluation model score and the pre-evaluation model.
The processor 1000 is further configured to read the program and perform the following steps:
and under the condition that the risk evaluation score is greater than or equal to a preset threshold value, determining that the user to be evaluated passes through risk evaluation.
As shown in fig. 11, an electronic device according to an embodiment of the present invention includes: the processor 1100, configured to read the program in the memory 1120, performs the following procedures:
Receiving a service request of a user;
acquiring user information according to the service request of the user;
Based on the user information, evaluating the user by using the evaluating method to obtain a risk evaluating score of the user;
And obtaining a business auditing result of the user according to the risk evaluating score.
A transceiver 1110 for receiving and transmitting data under the control of the processor 1100.
Wherein in fig. 11, a bus architecture may comprise any number of interconnected buses and bridges, and in particular one or more processors represented by processor 1100 and various circuits of memory represented by memory 1120, linked together. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, power management circuits, etc., which are well known in the art and, therefore, will not be described further herein. The bus interface provides an interface. The transceiver 1110 may be a number of elements, including a transmitter and a transceiver, providing a means for communicating with various other apparatus over a transmission medium. The processor 1100 is responsible for managing the bus architecture and general processing, and the memory 1120 may store data used by the processor 1100 in performing operations.
The processor 1100 is responsible for managing the bus architecture and general processing, and the memory 1120 may store data used by the processor 1100 in performing operations.
The embodiment of the invention also provides a computer readable storage medium, and a computer program is stored on the computer readable storage medium, and when the computer program is executed by a processor, the computer program realizes each process of the embodiment of the evaluation model training method or the evaluation method or the business auditing method, and can achieve the same technical effect, so that repetition is avoided, and the description is omitted here. The computer readable storage medium is, for example, a Read-Only Memory (ROM), a random access Memory (Random Access Memory RAM), a magnetic disk or an optical disk.
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 apparatus 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 apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. In light of such understanding, the technical solutions of the present invention may be embodied essentially or in part in the form of a software product stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) comprising instructions for causing a terminal (which may be a cell phone, computer, server, air conditioner, or network device, etc.) to perform the methods described in the various embodiments of the present invention.
The embodiments of the present invention have been described above with reference to the accompanying drawings, but the present invention is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those having ordinary skill in the art without departing from the spirit of the present invention and the scope of the claims, which are to be protected by the present invention.
Claims (11)
1. An evaluation model training method, comprising the steps of:
Acquiring sample data;
Inputting the sample data into a pre-evaluation model to obtain a pre-evaluation score of the sample data;
Extracting intermediate sample data of the sample data based on the pre-evaluation score, wherein the pre-evaluation score of the intermediate sample data is larger than a first preset value and smaller than a second preset value, the first preset value is smaller than the second preset value, and the first preset value and the second preset value are both larger than 0;
inputting the intermediate sample data into a neural network for model training to obtain a risk evaluation model of the intermediate sample data;
the sample data comprises a plurality of sub-sample data corresponding to a plurality of data sources, and the pre-evaluation model is constructed through the following steps:
Dividing sub-sample data of each data source into at least two guest group data, and constructing a sub-pre-evaluation model for each guest group data; the sub pre-evaluation model corresponding to each guest group data is constructed based on N sub-models corresponding to the guest group data and a first preset algorithm, one sub-model of each guest group data is constructed based on target data matched with a model target of the sub-model in the corresponding guest group data and a second preset algorithm, and N is an integer equal to or equal to 1;
Fusing the sub pre-evaluation models of the guest group data to obtain a pre-evaluation model, wherein the method comprises the following steps:
and obtaining the pre-evaluation model based on the output of the sub-pre-evaluation models of all the guest groups and a third preset algorithm, wherein the third preset algorithm comprises a linear model or a tree model.
2. The method of claim 1, wherein the sample data comprises sub-sample data of at least one data source;
The method further comprises the steps of: the pre-evaluation model is obtained, which specifically comprises the following steps:
dividing each sub-sample data into at least two guest group data respectively;
Obtaining a sub pre-evaluation model of each guest group data;
And fusing the sub pre-evaluation models of the guest group data to obtain the pre-evaluation model.
3. The method of claim 2, wherein the obtaining a sub-pre-evaluation model for each group of guests comprises:
Determining a sub-model corresponding to first guest group data, the first guest group data being any guest group data in all guest group data of all data sources;
And obtaining a sub pre-evaluation model of the first customer group data based on the sub model corresponding to the first customer group data and a first preset algorithm.
4. A method according to claim 3, wherein said determining a sub-model to which said first customer group data corresponds comprises:
selecting target data from the first customer base data; wherein the target data corresponds to a model target;
and obtaining a sub-model of the first customer group data based on the target data and a second preset algorithm.
5. A method according to claim 2 or 3, wherein the inputting the intermediate sample data into a neural network for model training, to obtain a risk assessment model of the intermediate sample data, comprises:
Determining target group data corresponding to the intermediate sample data from at least two group data of each data source;
obtaining a sub-pre-evaluation model of the target guest group data;
and fusing the sub pre-evaluation models of the target guest group data to obtain a risk evaluation model of the intermediate sample data.
6. An evaluation method, comprising:
acquiring user information of a user to be evaluated;
According to the user information, determining guest group information corresponding to the user to be evaluated;
Based on a pre-evaluation model corresponding to the guest group information, obtaining a pre-evaluation score of the user to be evaluated;
Inputting the user information into a risk evaluation model under the condition that the pre-evaluation score is larger than a first preset value and smaller than a second preset value to obtain the risk evaluation score of the user to be evaluated; the risk assessment model is obtained by the method of any one of claims 1-5;
wherein the first preset value and the second preset value are both greater than 0.
7. The method according to claim 6, wherein the obtaining the pre-evaluation score of the user to be evaluated based on the pre-evaluation model corresponding to the guest group information includes:
obtaining a sub-model score based on the sub-model corresponding to the user information and the guest group information;
Based on the sub-model score and a sub-pre-evaluation model corresponding to the guest group information, obtaining a sub-pre-evaluation model score corresponding to the user information;
And obtaining the pre-evaluation score of the user to be evaluated based on the sub-pre-evaluation model score and the pre-evaluation model.
8. A business auditing method, comprising:
Receiving a service request of a user;
acquiring user information according to the service request of the user;
Evaluating the user by using the evaluating method according to any one of claim 6 or claim 7 based on the user information to obtain a risk evaluating score of the user;
And obtaining a business auditing result of the user according to the risk evaluating score.
9. A business audit device, comprising:
The first receiving module is used for receiving the service request of the user;
The first acquisition module is used for acquiring user information according to the service request of the user;
A first evaluation module, configured to evaluate the user by using the evaluation method according to any one of claims 6 or 7 based on the user information, to obtain a risk evaluation score of the user;
and the second acquisition module is used for acquiring a business auditing result of the user according to the risk evaluation score.
10. An electronic device, comprising: a memory, a processor, and a program stored on the memory and executable on the processor; it is characterized in that the method comprises the steps of,
The processor for reading a program implementation in a memory comprising the steps of the method for training an evaluation model according to any one of claims 1 to 5; or to implement the steps in an evaluation method comprising any one of claims 6 to 7; or to implement the steps comprising the business audit method according to claim 8.
11. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the steps comprising the method for training an evaluation model according to any one of claims 1 to 5; or to implement the steps in an evaluation method comprising any one of claims 6 to 7; or to implement the steps comprising the business audit method according to claim 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010299090.4A CN113537666B (en) | 2020-04-16 | 2020-04-16 | Evaluation model training method, evaluation and business auditing method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010299090.4A CN113537666B (en) | 2020-04-16 | 2020-04-16 | Evaluation model training method, evaluation and business auditing method, device and equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113537666A CN113537666A (en) | 2021-10-22 |
| CN113537666B true CN113537666B (en) | 2024-05-03 |
Family
ID=78120212
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010299090.4A Active CN113537666B (en) | 2020-04-16 | 2020-04-16 | Evaluation model training method, evaluation and business auditing method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113537666B (en) |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104537389A (en) * | 2014-12-29 | 2015-04-22 | 生迪光电科技股份有限公司 | Human face recognition method and terminal equipment |
| CN105787437A (en) * | 2016-02-03 | 2016-07-20 | 东南大学 | Vehicle brand type identification method based on cascading integrated classifier |
| CN106352244A (en) * | 2016-08-31 | 2017-01-25 | 中国石油化工股份有限公司 | Pipeline leakage detection method based on hierarchical neural network |
| CN107040397A (en) * | 2016-02-03 | 2017-08-11 | 腾讯科技（深圳）有限公司 | A kind of service parameter acquisition methods and device |
| CN107133743A (en) * | 2017-05-19 | 2017-09-05 | 重庆文理学院 | A kind of appraisal procedure of small micro- enterprise operation project |
| EP3276540A2 (en) * | 2016-07-28 | 2018-01-31 | Samsung Electronics Co., Ltd. | Neural network method and apparatus |
| CN107945003A (en) * | 2016-12-30 | 2018-04-20 | 上海壹账通金融科技有限公司 | Credit estimation method and device |
| CN108053310A (en) * | 2017-11-24 | 2018-05-18 | 深圳市牛鼎丰科技有限公司 | Credit scoring method, apparatus, computer equipment and storage medium |
| CN109816509A (en) * | 2018-12-14 | 2019-05-28 | 平安科技（深圳）有限公司 | Generation method, terminal device and the medium of scorecard model |
| CN110097450A (en) * | 2019-03-26 | 2019-08-06 | 中国人民财产保险股份有限公司 | Vehicle borrows methods of risk assessment, device, equipment and storage medium |
| 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 |
| CN110264274A (en) * | 2019-06-21 | 2019-09-20 | 深圳前海微众银行股份有限公司 | Objective group's division methods, model generating method, device, equipment and storage medium |
| WO2019180310A1 (en) * | 2018-03-21 | 2019-09-26 | Nokia Technologies Oy | A method, an apparatus and a computer program product for an interpretable neural network representation |
| CN110400215A (en) * | 2019-07-31 | 2019-11-01 | 浪潮软件集团有限公司 | Small micro- Enterprise Credit Rating Model construction method and system towards family, enterprise |
| CN110458690A (en) * | 2019-07-30 | 2019-11-15 | 联动优势科技有限公司 | A kind of method for building up and device of credit scoring model |
| CN110717650A (en) * | 2019-09-06 | 2020-01-21 | 平安医疗健康管理股份有限公司 | Receipt data processing method and device, computer equipment and storage medium |
| CN110956273A (en) * | 2019-11-07 | 2020-04-03 | 中信银行股份有限公司 | Credit scoring method and system integrating multiple machine learning models |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6847386B2 (en) * | 2016-09-09 | 2021-03-24 | インターナショナル・ビジネス・マシーンズ・コーポレーションＩｎｔｅｒｎａｔｉｏｎａｌ Ｂｕｓｉｎｅｓｓ Ｍａｃｈｉｎｅｓ Ｃｏｒｐｏｒａｔｉｏｎ | Neural network regularization |
| US11487941B2 (en) * | 2018-05-21 | 2022-11-01 | State Street Corporation | Techniques for determining categorized text |
| US11755975B2 (en) * | 2018-08-10 | 2023-09-12 | Visa International Service Association | System, method, and computer program product for implementing a hybrid deep neural network model to determine a market strategy |
-
2020
- 2020-04-16 CN CN202010299090.4A patent/CN113537666B/en active Active
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104537389A (en) * | 2014-12-29 | 2015-04-22 | 生迪光电科技股份有限公司 | Human face recognition method and terminal equipment |
| CN105787437A (en) * | 2016-02-03 | 2016-07-20 | 东南大学 | Vehicle brand type identification method based on cascading integrated classifier |
| CN107040397A (en) * | 2016-02-03 | 2017-08-11 | 腾讯科技（深圳）有限公司 | A kind of service parameter acquisition methods and device |
| EP3276540A2 (en) * | 2016-07-28 | 2018-01-31 | Samsung Electronics Co., Ltd. | Neural network method and apparatus |
| CN106352244A (en) * | 2016-08-31 | 2017-01-25 | 中国石油化工股份有限公司 | Pipeline leakage detection method based on hierarchical neural network |
| CN107945003A (en) * | 2016-12-30 | 2018-04-20 | 上海壹账通金融科技有限公司 | Credit estimation method and device |
| CN107133743A (en) * | 2017-05-19 | 2017-09-05 | 重庆文理学院 | A kind of appraisal procedure of small micro- enterprise operation project |
| CN108053310A (en) * | 2017-11-24 | 2018-05-18 | 深圳市牛鼎丰科技有限公司 | Credit scoring method, apparatus, computer equipment and storage medium |
| WO2019180310A1 (en) * | 2018-03-21 | 2019-09-26 | Nokia Technologies Oy | A method, an apparatus and a computer program product for an interpretable neural network representation |
| CN109816509A (en) * | 2018-12-14 | 2019-05-28 | 平安科技（深圳）有限公司 | Generation method, terminal device and the medium of scorecard model |
| CN110097450A (en) * | 2019-03-26 | 2019-08-06 | 中国人民财产保险股份有限公司 | Vehicle borrows methods of risk assessment, device, equipment and storage medium |
| 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 |
| CN110264274A (en) * | 2019-06-21 | 2019-09-20 | 深圳前海微众银行股份有限公司 | Objective group's division methods, model generating method, device, equipment and storage medium |
| CN110458690A (en) * | 2019-07-30 | 2019-11-15 | 联动优势科技有限公司 | A kind of method for building up and device of credit scoring model |
| CN110400215A (en) * | 2019-07-31 | 2019-11-01 | 浪潮软件集团有限公司 | Small micro- Enterprise Credit Rating Model construction method and system towards family, enterprise |
| CN110717650A (en) * | 2019-09-06 | 2020-01-21 | 平安医疗健康管理股份有限公司 | Receipt data processing method and device, computer equipment and storage medium |
| CN110956273A (en) * | 2019-11-07 | 2020-04-03 | 中信银行股份有限公司 | Credit scoring method and system integrating multiple machine learning models |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113537666A (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106875078B (en) | Transaction risk detection method, device and equipment | |
| Dinh et al. | A credit scoring model for Vietnam's retail banking market | |
| CN111291816B (en) | Method and device for carrying out feature processing aiming at user classification model | |
| CN109583966B (en) | High-value customer identification method, system, equipment and storage medium | |
| CN112734559B (en) | Enterprise credit risk evaluation method and device and electronic equipment | |
| CN108876188B (en) | Inter-connected service provider risk assessment method and device | |
| CN104321794A (en) | A system and method using multi-dimensional rating to determine an entity's future commercial viability | |
| CN110930218B (en) | Method and device for identifying fraudulent clients and electronic equipment | |
| RU2640633C2 (en) | Calculation of probability that company complies with its obligations | |
| CN110728301A (en) | Credit scoring method, device, terminal and storage medium for individual user | |
| CN112232950A (en) | Loan risk assessment method and device, equipment and computer-readable storage medium | |
| KR101999765B1 (en) | Method and system for analyzing pattern of banking phishing loan using data mining technology | |
| CN114066606A (en) | System and method for falsely identifying data based on text escape as GPS distance | |
| CN109146667B (en) | Method for constructing external interface comprehensive application model based on quantitative statistics | |
| CN115545886A (en) | Overdue risk identification method, overdue risk identification device, overdue risk identification equipment and storage medium | |
| CN112950359B (en) | User identification method and device | |
| CN114066620A (en) | Client information processing method and device based on client portrait | |
| CN113537666B (en) | Evaluation model training method, evaluation and business auditing method, device and equipment | |
| CN114066478A (en) | Target classification evaluation method and device based on informed production information | |
| CN110570301B (en) | Risk identification method, device, equipment and medium | |
| CN114240605A (en) | Loan calculation method and device, computer equipment and storage medium | |
| CN112927069A (en) | User credit risk assessment method and device and electronic equipment | |
| CN113313582A (en) | Guest refusing and reflashing model training method and device and electronic equipment | |
| CN111461863A (en) | Data processing method and device, computer equipment and storage medium | |
| CN110956471A (en) | Method for analyzing credit investigation data of decoration industry |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |