CN114372861A

CN114372861A - Data processing method, data processing device, computer equipment and storage medium

Info

Publication number: CN114372861A
Application number: CN202111489397.1A
Authority: CN
Inventors: 颜书云; 刘庆
Original assignee: Nanjing Xingyun Digital Technology Co Ltd
Current assignee: Nanjing Xingyun Digital Technology Co Ltd
Priority date: 2021-12-08
Filing date: 2021-12-08
Publication date: 2022-04-19

Abstract

The application relates to a data processing method, a data processing device, computer equipment and a storage medium. The method comprises the following steps: acquiring resource borrowing application data of a current user; acquiring a current resource lending type, resource lending duration corresponding to the current resource lending type and resource transaction behavior data between the current user and each resource lending institution in the resource lending duration according to the resource lending application data; obtaining the credit score of the current user according to a preset risk assessment model; determining a current risk control range; and when the credit score is within the current risk control range, determining that the resource borrowing application data of the current user passes the audit. According to the resource borrowing method and device, the current resource borrowing type, the corresponding resource borrowing duration and the resource transaction behavior data are combined with the credit score of the user, the resource borrowing application data are subjected to refining, the refining degree of a data source is improved, and the auditing result is more accurate.

Description

Data processing method, data processing device, computer equipment and storage medium

Technical Field

The present application relates to the field of data processing technologies, and in particular, to a data processing method and apparatus, a computer device, and a storage medium.

Background

At present, each consumption financial institution, the credit admission risk strategy mainly adopts top-level subdivision portrayal based on the attributes of customer acquisition channels, products and demographic characteristics: channels, products, and demographics Attributes. The credit consumption field usually takes a passenger obtaining channel or a product as a main upper dimension and takes demographic characteristics (age, academic calendar and occupation) as a lower dimension to construct a credit granting access top-level subdivision portrayal framework.

However, in the conventional technical solution, the group characterization based on the population characteristics is limited by the unbalanced influence of the organization on the learning credit network and the acquisition of the professional information, the data loss and the passenger service fairness (such as gender), so that the application of the subdivision characterization in the credit passenger group application trust phase is limited, and the dimension of the population characteristics adopted by the conventional technique is not fine enough and single, which easily results in the credit application verification according to the conventional scheme, and the verification accuracy is not high.

Disclosure of Invention

In view of the above, it is necessary to provide a data processing method, an apparatus, a computer device and a storage medium for solving the above technical problems.

A method of data processing, the method comprising:

acquiring resource borrowing application data of a current user;

acquiring a current resource lending type, a resource lending duration corresponding to the current resource lending type and resource transaction behavior data between a current user and each resource lending institution within the resource lending duration according to the resource lending application data;

obtaining a credit score of the current user according to a preset risk assessment model;

acquiring a current risk control range, wherein the current risk control range is a preset risk control range corresponding to a current resource borrowing type, a resource borrowing duration and resource transaction behavior data;

and when the credit score is within the current risk control range, determining that the resource borrowing application data of the current user is approved.

In one embodiment, the method further includes:

acquiring sample data, wherein the sample data comprises resource borrowing application data corresponding to a preset number of sample users;

performing model training according to the sample data and the credit scores of the sample users to obtain a decision tree model, wherein the decision tree model comprises a first level, a second level and a third level, each node in the first level is configured with a corresponding resource borrowing type, each node in the second level is configured with a duration condition corresponding to each resource borrowing type, and each node in the third level is configured with a corresponding risk control range and a resource transaction condition under each duration condition;

the obtaining of the current risk control range includes:

matching the resource lending duration with each duration condition under the current resource lending type, and determining the duration condition matched with the resource lending duration;

matching the resource transaction behavior data with each resource transaction condition under the time length condition matched with the resource lending time length, and determining the resource transaction condition matched with the resource transaction behavior data;

and acquiring a risk control range configured on the node corresponding to the resource transaction condition matched with the resource transaction behavior data to obtain a current risk control range.

In one embodiment, the method further includes:

obtaining default rate and audit passing rate of each node in the decision tree model;

determining the average default rate and the average audit passing rate of each node;

comparing the default rate of each node in the decision tree model with the average default rate to obtain a high default rate node and a low default rate node;

comparing the audit passing rate of each node in the decision tree model with the average audit passing rate to obtain high-passing-rate nodes and low-passing-rate nodes;

when a node with a high default rate and a high passing rate or a node with a low default rate and a low passing rate is screened out from all nodes, new credit scores of the sample users corresponding to the node with the high default rate and the high passing rate or the node with the low default rate and the low passing rate are obtained again;

and retraining the decision tree model according to the new credit score.

In one embodiment, the method further includes:

and when the decision tree model after retraining has high default rate high pass rate nodes or low default rate low pass rate nodes, adjusting the risk control ranges corresponding to the high default rate low pass rate nodes and the low default rate high pass rate nodes according to the default rate of the high default rate high pass rate nodes and the audit pass rate, the default rate of the low default rate low pass rate nodes and the audit pass rate.

In one embodiment, the method further includes:

training sample data to obtain a rejection sample and a pass sample;

fishing a corresponding number of rejected samples from the rejected samples according to a preset input ratio, and retraining the fished rejected samples;

removing a corresponding number of passing samples from the passing samples according to a preset set-out ratio to update the passing samples;

the obtaining of the audit passing rate of each node in the decision tree model includes:

and determining the auditing passing rate of each node in the decision tree model according to the updated number of the passing samples.

In one embodiment, the method further includes:

receiving a request of a new resource transfer type in a second level;

and extracting the target type in the request of the newly added resource borrowing type, and adding the target type into the decision tree model.

In one embodiment, the obtaining the current resource lending type, the resource lending duration corresponding to the current resource lending type, and the resource transaction behavior data between the current user and each resource lending mechanism in the resource lending duration according to the resource lending application data includes:

extracting a current resource borrowing type from the resource borrowing application data;

acquiring historical resource lending data of a current user according to the current resource lending type and user information in the resource lending application data;

acquiring the duration from the time when the current user applies for the resource corresponding to the current resource transfer type for the first time to the current time according to the historical resource transfer data;

and acquiring resource transaction behavior data between the current user and each resource lending mechanism from the time when the current user applies for the resource corresponding to the current resource lending type for the first time to the current time according to the historical resource lending data.

A data processing apparatus, characterized in that the apparatus comprises:

the first acquisition module is used for acquiring resource borrowing application data of a current user;

the second acquisition module is used for acquiring the current resource lending type, the resource lending duration corresponding to the current resource lending type and the resource transaction behavior data between the current user and each resource lending institution within the resource lending duration according to the resource lending application data;

the generating module is used for obtaining the credit score of the current user according to a preset risk evaluation model;

the determining module is used for acquiring a current risk control range, wherein the current risk control range is a preset risk control range corresponding to the current resource transfer type, the resource transfer duration and the resource transaction behavior data;

and the auditing module is used for determining that the resource borrowing application data of the current user is approved when the credit score is within the current risk control range.

A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:

acquiring resource borrowing application data of a current user;

A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:

acquiring resource borrowing application data of a current user;

The data processing method, the data processing device, the computer equipment and the storage medium acquire the resource borrowing application data of the current user; acquiring a current resource lending type, a resource lending duration corresponding to the current resource lending type and resource transaction behavior data between a current user and each resource lending institution within the resource lending duration according to the resource lending application data; obtaining a credit score of the current user according to a preset risk assessment model; acquiring a current risk control range, wherein the current risk control range is a preset risk control range corresponding to a current resource borrowing type, a resource borrowing duration and resource transaction behavior data; and when the credit score is within the current risk control range, determining that the resource borrowing application data of the current user is approved. According to the resource borrowing method and device, the current resource borrowing type, the corresponding resource borrowing duration, the resource transaction behavior data and the credit score of the user are combined, the refinement degree of a data source is improved for the resource borrowing application data, the auditing result is more accurate, and the population characteristic grouping portrayal is not used singly in the method, so that the influence of the imbalance of user information acquisition is avoided.

Drawings

FIG. 1 is a diagram of an application environment of a data processing method in one embodiment;

FIG. 2 is a flow diagram illustrating a data processing method according to one embodiment;

FIG. 3 is a diagram illustrating the structure of an adaptation decision tree model according to an embodiment;

FIG. 4 is a flow diagram illustrating tuning a decision tree model according to one embodiment;

fig. 5 is a flowchart illustrating a detailed process of the step of acquiring the current resource lending type, the resource lending duration corresponding to the current resource lending type, and the resource transaction behavior data between the current user and each resource lending institution within the resource lending duration according to the resource lending application data in another embodiment;

FIG. 6 is a block diagram showing the structure of a data processing apparatus according to an embodiment;

FIG. 7 is a diagram illustrating an internal structure of a computer device according to an embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

Referring to fig. 1, fig. 1 is a schematic diagram of an application environment of a data processing method according to an exemplary embodiment of the present application. As shown in fig. 1, the application environment includes a server 100 and a terminal 101, and the server 100 and the terminal 101 can be communicatively connected through a network 102 to implement the data processing method of the present application.

The server 100 is configured to obtain resource lending application data of a current user; acquiring a current resource lending type, a resource lending duration corresponding to the current resource lending type and resource transaction behavior data between a current user and each resource lending institution within the resource lending duration according to the resource lending application data; obtaining a credit score of the current user according to a preset risk assessment model; acquiring a current risk control range, wherein the current risk control range is a preset risk control range corresponding to a current resource borrowing type, a resource borrowing duration and resource transaction behavior data; and when the credit score is within the current risk control range, determining that the resource borrowing application data of the current user passes the audit, receiving an inquiry request of the audit result sent by the terminal 101, obtaining the corresponding audit result, and sending the audit result to the terminal for display. The server 100 may be implemented as an independent server or a server cluster composed of a plurality of servers.

The terminal 101 is configured to send a query request of the audit result to the server 100, receive the audit result fed back by the server 100, and display the audit result. The terminal 101 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices.

The network 102 is used for network connection between the terminal 101 and the server 100, and in particular, the network 102 may include various types of wired or wireless networks.

In one embodiment, as shown in fig. 2, a data processing method is provided, which is described by taking the application of the method to the server in fig. 1 as an example, and includes the following steps:

and S11, acquiring the resource borrowing application data of the current user.

In the present application, the resource lending application data is data related to the resource lending application, such as user information of the application user, a resource lending type, a resource lending duration, resource transaction behavior data, a resource lending amount, and the like. In an application scenario, the resource lending application data may be credit application data. The resource transfer type may be a credit type, the resource transfer duration may be a credit duration, that is, a duration from the user's account to the present, and the resource transaction behavior data may be multi-head data, that is, a number of institutions by which the user initiated a behavior related to credit within the duration from the user's account to the present.

Specifically, the credit types may be a minor loan type, a cash payment type, a bank individual loan type, a credit card type, other loan types, a weak pedestrian group, an unmanned record group, and the like. The credit duration under each credit type respectively corresponds to the duration of the first account opening to the present under different credit types.

And S12, acquiring the current resource lending type, the resource lending duration corresponding to the current resource lending type and the resource transaction behavior data between the current user and each resource lending institution within the resource lending duration according to the resource lending application data.

In the present application, the resource lending institution may be an institution, such as a credit institution, that lends the resource to the applicant. The resource transaction behavior data may be the mechanism number of the resource transaction behavior initiated by the current user within the resource lending duration, for example, the resource transaction behavior data may be multi-head data.

For example, the current user applies for a loan to the institution a, initiates a loan application, and the institution a obtains the credit type applied by the current user and the loan duration since the first account opening of the current user, and the number of institutions in which the current user initiates credit with other institutions.

In one embodiment, as shown in fig. 3, the obtaining of the current resource lending type, the resource lending duration corresponding to the current resource lending type, and the resource transaction behavior data between the current user and each resource lending institution within the resource lending duration according to the resource lending application data may include:

s51, extracting the current resource borrowing type from the resource borrowing application data;

s52, acquiring historical resource lending data of the current user according to the current resource lending type and the user information in the resource lending application data;

s53, acquiring the duration from the time when the current user applies for the resource corresponding to the current resource transfer type for the first time to the current time according to the historical resource transfer data;

and S54, acquiring resource transaction behavior data between the current user and each resource lending mechanism from the time when the current user applies for the resource corresponding to the current resource lending type for the first time to the current time according to the historical resource lending data.

In this application, the resource transaction behavior data may be the number of resource lending institutions for which the resource lending duration is the resource application-related behavior initiated by the current user. For example, if the resource transfer duration is 3 months, the number of credit institutions for which the current user applied for loan in the last 3 months, or the number of credit institutions for which an action related to the credit application was initiated, is obtained. The credit-related behavior may be a credit application, credit, and repayment behavior.

According to the method and the device, the current resource borrowing type, the resource borrowing duration and the resource transaction behavior data are obtained, the resource borrowing application data of the user are checked based on the data, the user resource borrowing application data are checked by the method and the device, the user credit is actually checked, the default risk of the user passing the check is low as much as possible, and the accuracy of the check is improved.

And S13, obtaining the credit score of the current user according to a preset risk assessment model.

In this application, the risk assessment model may be used to score the credit rating of the user according to the historical resource lending data of the user. For example, the credit score of the user may be finally obtained according to information such as loan type, loan institution, default condition, repayment condition, promissory condition, overdue condition and the like applied by the user history as input parameters.

And S14, acquiring a current risk control range, wherein the current risk control range is a preset risk control range corresponding to the current resource borrowing type, the resource borrowing duration and the resource transaction behavior data.

In the method and the system, a pre-configured risk control range can be uniquely determined according to the resource transaction behavior data, the current resource lending type and the corresponding resource lending duration. Specifically, the risk control range is configured through a decision tree model.

And S15, when the credit score is in the current risk control range, determining that the resource borrowing application data of the current user is approved.

In one embodiment, the method may further include:

the obtaining of the current risk control range may include:

Referring to fig. 4, fig. 4 is a schematic diagram of a model structure of a decision tree according to an embodiment. In fig. 4, the decision tree model includes a plurality of levels, each level includes a plurality of nodes, and each node is configured with a classification condition. The classification condition configured on each node on the first hierarchy level 32 is a resource lending type, wherein the nodes on the first hierarchy level 32 respectively include 321, 323 and other nodes (not shown). The classification condition configured on each node on the second hierarchy 33 is a duration condition, and the duration condition is a condition satisfied by the resource lending duration, wherein the nodes on the second hierarchy 33 respectively include 331, 332, 333 and other nodes (not shown). The classification condition configured on each node on the third hierarchy 34 is a resource transaction condition, and the resource transaction condition is a condition satisfied by the resource transaction data, wherein the nodes on the third hierarchy 34 respectively include 341, 342 and other nodes (not shown). In the present application, a corresponding risk control range is also configured for each node on the third hierarchy level 34. Further, the decision tree model of the present application further includes a base level 31, where the classification condition of the base level may be a channel or a product, where the base level includes a node 311 and other nodes (not shown). Different preset channels or preset products are distributed and configured on each node of the basic level. In another embodiment, the base hierarchy may include a plurality of sub-hierarchies, for example, a first sub-hierarchy is configured as a product, and a second sub-hierarchy is configured as a channel.

Specifically, the decision tree further includes a fourth hierarchy 35, wherein nodes in the fourth hierarchy 35 are 351 and 352, respectively, and other nodes (not shown).

When training sample data, acquiring current sample data which is resource borrowing application data of a current sample user. And inputting the resource borrowing application data into an initial model, and training sample data by using the initial model to obtain a final decision tree model.

Specifically, an application channel in current sample data is matched with a preset channel configured on each node in a basic level in an initial model, when the matching is successful, the current sample data is classified to an A node which is successfully matched in the basic level, the A node is assumed to be a low credit type node, and then the classification of the A node continues to the lower layer until the A node is classified to a certain M node of the last layer.

Further, the next level of the base level 31 is a first level 32, and each node of the first level 32 is configured with a corresponding resource lending type. For example, the second level includes 7 nodes, and the respective allocated resource lending types are a loan minor type, a fund consumption type, a bank individual lending type, a credit card type, other loan types, a weak pedestrian or an unmanned group type, where the resource lending type allocated on the node 321 is the loan minor type, and the resource lending type allocated on the node 322 is the weak pedestrian or the unmanned pedestrian. The current sample data classified to the node a of the first hierarchy 32 may be further classified to the node B at the second hierarchy 33 according to the current resource lending type.

Further, the next level of the first level 32 is a second level 33, and each node of the second level 33 is configured with a corresponding classification condition, which is a duration condition corresponding to each resource lending type. For example, the loan minor type is divided into 3 time length conditions, where the loan minor length is smaller than X, corresponding to node 331, the loan minor length is greater than X and smaller than Y, corresponding to node 332, the loan minor length is greater than Y, and corresponding to node 333. And respectively matching the resource lending duration in the current sample data with the three duration conditions under the loan type, and if the loan duration is matched to be less than the node X, assuming that the node C is the node C.

Further, the next level of the second level 33 is a third level 34, and the classification condition of each node length of the third level 34 is a resource transaction condition. The resource transaction condition may be high multi-headed when the resource transaction data is greater than a first preset threshold, corresponding to node 341, and low multi-headed when the resource transaction data is less than a second preset threshold, corresponding to node 342. And matching the resource lending duration in the current sample data on the node C with the resource transaction condition in the third level, and if the node with the highest head is matched, assuming that the node is a node D, acquiring a risk control range configured in advance on the node D. In the present application, a corresponding risk control range is also configured in advance for each node in the third level 34.

Further, a credit score of a current sample user is obtained according to current sample data, the credit score is compared with a risk control range on the D node, if the credit score is within the risk control range on the D node, it is determined that the current sample data passes the audit, and then the next sample data is continuously obtained for the audit until the last sample data, so as to train the decision tree model.

In the method, the channel, the credit type and the credit duration are selected as the upper layer, the multi-head information and the credit score are selected as the lower layer, and the obtained data are refined, so that the auditing result of the trained decision tree model is more accurate when the credit application data is audited.

In one embodiment, as shown in fig. 5, the method further includes:

s41, obtaining default rate and audit passing rate of each node in the decision tree model;

s42, determining the average default rate and the average audit passing rate of each node;

s43, comparing the default rate of each node in the decision tree model with the average default rate to obtain a high default rate node and a low default rate node;

s44, comparing the audit passing rate of each node in the decision tree model with the average audit passing rate to obtain a high-passing-rate node and a low-passing-rate node;

s45, when the nodes with high default rate and high passing rate or the nodes with low default rate and low passing rate are screened out from the nodes, new credit scores of the sample users corresponding to the nodes with high default rate and high passing rate or the nodes with low default rate and low passing rate are obtained again;

and S46, retraining the decision tree model according to the new credit score.

In this application, the obtaining of the default rate and the audit passing rate of each node in the decision tree model can be implemented by the following steps:

acquiring the number of sample users distributed to each node in the third level and the number of sample users passing the audit, and acquiring the audit passing rate of each node in the third level according to the number of distributed sample users and the number of sample users passing the audit;

the default data of the sample users distributed to each node in the third hierarchy is obtained, specifically, the number of the default sample users is obtained, and the default rate of each node in the third hierarchy is obtained according to the number of the default sample users and the number of the sample users distributed to each node.

The above comparing the default rate of each node in the decision tree model with the average default rate to obtain the high default rate node and the low default rate node can be realized by the following steps:

determining that the default rate is smaller than a third preset threshold as a low default rate, and determining that the default rate is larger than a fourth preset threshold as a high default rate;

and determining that the audit passing rate is smaller than a fifth preset threshold as a low passing rate, and determining that the audit passing rate is larger than a sixth preset threshold as a high passing rate.

The third preset threshold may be 50% of the average default rate. The fourth preset threshold may be 15% of the average default rate. The fifth preset threshold may be 50% of the average passing rate. The sixth preset threshold may be 1.5 times of the average passing rate. In the present application, the specific preset threshold may also be specifically set according to actual requirements, and is not specifically limited herein.

When the nodes with high default rate and high passing rate or the nodes with low default rate and low passing rate are screened out from the nodes, the new credit scores of the nodes with high default rate and high passing rate or the nodes with low default rate and low passing rate corresponding to the sample users are obtained again, and the new credit scores are replaced with the original scores to train the sample users again.

In the present application, the training process of the decision tree model is a process of continuously adjusting the initial model. Specifically, when model training is performed on sample data, parameters of the decision tree model are continuously adjusted according to the default rate and the audit passing rate of each node to obtain an optimal decision tree model, wherein the parameters of the decision tree model can be the risk control range.

In one embodiment, the method may further include:

In the application, when a decision tree model after retraining can still screen out nodes with high default rate and high passing rate or nodes with low default rate and low passing rate, according to the default rate and the audit passing rate of the nodes with high default rate and high passing rate, the default rate and the audit passing rate of the nodes with low default rate and low passing rate, the risk control ranges corresponding to the nodes with high default rate and low default rate and high passing rate are adjusted, specifically, the risk control range of the nodes with high default rate and high passing rate is adjusted to be a smaller range, the risk control range of the nodes with low default rate and low passing rate is adjusted to be a larger range, the audit passing rate of the nodes with high default rate and high passing rate is reduced, and the passing rate of the nodes with low default rate and low passing rate is improved.

In the application, the decision tree model can be adjusted by improving the implementation mode, so that the decision tree model is more accurate, and the flexibility of the scheme is improved.

In one embodiment, the method may further include:

training sample data to obtain a rejection sample and a pass sample;

obtaining the audit passing rate of each node in the decision tree model, including:

In this application, can put into the ratio to the low throughput rate node setting of high default rate to drag for the rejection sample of corresponding number back from rejecting the sample, train the rejection sample of dragging for again. A set-out ratio may be set for low default rates and high pass rates to cull a corresponding number of pass samples from the pass samples to update the pass samples.

In the application, a random replacement rate is usually set according to the degree of policy preference, under a loose scene, an insertion rate is not lower than an extraction rate value (namely a positive salvage rate), a 2.0-4.0% extraction rate is usually set for a node with low default rate and high passing rate, and a 3.0-5.0% insertion rate is set for a node with high default rate and low passing rate; in the tightening scenario, the placement ratio is not higher than the placement ratio, and typically, the placement ratio is set to 1.0% -3.0% for low default rate and low pass rate nodes, and the placement ratio is set to 1.0% -3.0% for high default rate and low pass rate nodes.

This application can compensate decision tree model through this embodiment, returns to dragging for partly in the sample that promptly refuses to fall, refuses partly again in the sample that passes through to the precision of adjustment model.

In one embodiment, the method may further include:

receiving a request of a new resource transfer type in a second level;

In the application, when a new service type needs to be added, a request for adding a resource transfer type can be submitted to the server. For example, when a new credit type is needed that is not available in the original decision tree model, the new credit type may be added to the decision tree model. When a new credit type is added to the decision tree model, the classification conditions configured on each node corresponding to the lower level of the new credit type need to be added to the decision tree model. For example, the duration condition for the new credit type and the resource transaction condition for each duration condition need to be configured.

In the application, the classification conditions of each node in each level in the decision tree model can be added or modified through the implementation mode, and the flexibility of the scheme is improved.

In a possible application scenario, a base level, a first level and a second level in the application are upper levels, a third level and a fourth level are lower levels, a decision tree model is mainly adopted during model training, a proper lower branch characteristic dimension is selected, lower branches grow layer by layer according to the ordering of decision tree characteristic chi-square statistics (F statistics or entropy) and upper-layer experience subdivision characteristics, and application risk scores are always used as the lowest-layer subdivision, and the method specifically comprises the following steps:

setting common parameters of the algorithm, wherein the significance level is 0.2, the maximum branch of the tree is 5, the maximum depth of the tree is 8, the minimum sample capacity of leaf nodes is 200, and other parameters are used for checking out default values;

the upper layer is divided into 6 groups (corresponding to the channels A-F) as shown in FIG. 4;

the lower subdivision of each upper-layer subdivision node grows, taking a channel A as an example, and the lower-layer subdivision queries the number of non-silver institutions of nearly 6 months as growth segmentation nodes by using the loan duration of a small loan institution and the multi-head loan application of a third party respectively;

applying for risk score (a-score) as a basedon/endnode feature selection and determining cutoff at different levels of overdue as the final split;

testing a forward bailing rule of the high-risk node, namely setting a certain putting ratio to carry out bailing on the rejected sample;

the low-risk node negative rejection rule test is carried out, namely a certain set-out ratio is set to reject the passing sample;

according to the steps, the growth of the lower branch decision tree of different channels and the research and development of forward and reverse cross tests are carried out, so that a final decision tree is formed;

after the model training is completed, the model needs to be evaluated. In the stage, the trained tree model risk default level under each upper-layer subdivision node is consistent, the stability of the subdivided passenger groups is evaluated, and then the risk homogeneity of the passenger groups in the same pool and the risk heterogeneity between the pools are tested (wherein the PSI deviation index of the proportion of the passenger groups is used for evaluating the stability of the passenger groups, and the mean T-test statistic of the default rate is used for the homogeneity of each branch).

And further performing application inspection, wherein in the stage, a model is mainly deployed on a policy engine, actual admission decision is performed on a new subdivision frame when 30% -50% of flow is applied randomly, decision results under different subdivision frames of the same batch of passenger groups are marked, and comparison and analysis are performed 120-180 days after credit admission so as to determine decision efficiency difference under different subdivision frames.

In conclusion, compared with the conventional scheme, the data processing method is relatively optimized in the aspects of the passing rate and the default rate, and the improvement of the model performance means that the distinguishing capability of the model on the credit risk is enhanced. Under the same business mode, by adopting the method, the credit approval passing rate is improved, the overdue credit risk is reduced, and a better risk control level is obtained.

In one embodiment, as shown in fig. 6, there is provided a data processing apparatus including: a first obtaining module 11, a second obtaining module 12, a generating module 13, a determining module 14 and an auditing module 15, wherein:

a first obtaining module 11, configured to obtain resource borrowing application data of a current user;

a second obtaining module 12, configured to obtain, according to the resource lending application data, a current resource lending type, a resource lending duration corresponding to the current resource lending type, and resource transaction behavior data between the current user and each resource lending institution within the resource lending duration;

the generating module 13 is configured to obtain a credit score of the current user according to a preset risk assessment model;

the determining module 14 is configured to obtain a current risk control range, where the current risk control range is a risk control range that is configured in advance and corresponds to a current resource lending type, a resource lending duration, and resource transaction behavior data;

and the auditing module 15 is used for determining that the resource borrowing application data of the current user is approved when the credit score is within the current risk control range.

In one embodiment, the apparatus further includes a training module (not shown), where the training module may obtain sample data, where the sample data includes resource lending application data corresponding to a preset number of sample users, and performs model training according to each sample data and credit score of each sample user to obtain a decision tree model, where the decision tree model includes a first level, a second level, and a third level, each node in the first level is configured with a corresponding resource lending type, each node in the second level is configured with a duration condition corresponding to each resource lending type, and each node in the third level is configured with a corresponding risk control range and a resource transaction condition under each duration condition;

the determining module 14 may match the resource lending duration with each duration condition in the current resource lending type, determine a duration condition matched with the resource lending duration, match the resource transaction behavior data with each resource transaction condition in the duration condition matched with the resource lending duration, determine a resource transaction condition matched with the resource transaction behavior data, obtain a risk control range configured on a node corresponding to the resource transaction condition matched with the resource transaction behavior data, and obtain a current risk control range.

In one embodiment, the apparatus further includes a first adjusting module (not shown), the first adjusting module can obtain the default rate and the audit passing rate of each node in the decision tree model, determine the average default rate and the average audit passing rate of each node, compare the default rate and the average default rate of each node in the decision tree model to obtain a high default rate node and a low default rate node, compare the audit passing rate and the average audit passing rate of each node in the decision tree model to obtain a high passing rate node and a low passing rate node, when the nodes with high default rate and high passing rate or the nodes with low default rate and low passing rate are screened out from the nodes, and re-acquiring a new credit score of the sample user corresponding to the high default rate and high pass rate node or the low default rate and low pass rate node, and re-training the decision tree model according to the new credit score.

In one embodiment, the apparatus further includes a second adjusting module (not shown), and when there is a high-default-rate high-throughput node or a low-default-rate low-throughput node in the re-trained decision tree model, the second adjusting module can adjust the risk control ranges corresponding to the high-default-rate low-throughput node and the low-default-rate high-throughput node according to the default rate and the audit throughput of the high-default-rate high-throughput node, the default rate and the audit throughput of the low-default-rate low-throughput node, and the audit throughput.

In one embodiment, the apparatus further includes a third adjusting module (not shown), where the third adjusting module is capable of training the sample data to obtain reject samples and pass samples, recalling a corresponding number of reject samples from the reject samples according to a preset input ratio, retraining the recalled reject samples, and eliminating a corresponding number of pass samples from the pass samples according to a preset output ratio to update the pass samples;

the first adjusting module may further determine an audit passing rate of each node in the decision tree model according to the updated number of passing samples.

In one embodiment, the apparatus further includes a newly adding module (not shown), and the newly adding module may receive a request for a newly added resource lending type in the second hierarchy, extract a target type in the request for the newly added resource lending type, and add the target type to the decision tree model.

In one embodiment, the resource lending duration is a duration from a time when the current user first applies for the resource corresponding to the current resource lending type to the current time, the second obtaining module 12 may extract the current resource lending type from the resource lending application data, obtain historical resource lending data of the current user according to the current resource lending type and user information in the resource lending application data, obtain a duration from a time when the current user first applies for the resource corresponding to the current resource lending type to the current time according to the historical resource lending data, and obtain resource transaction behavior data between the current user and each resource lending institution from the time when the current user first applies for the resource corresponding to the current resource lending type to the current time according to the historical resource lending data.

In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 7. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer equipment is used for storing data such as operation data of the intelligent household equipment. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a resource allocation method of a compiled virtual machine.

In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program: acquiring resource borrowing application data of a current user; acquiring a current resource lending type, a resource lending duration corresponding to the current resource lending type and resource transaction behavior data between a current user and each resource lending institution within the resource lending duration according to the resource lending application data; obtaining a credit score of the current user according to a preset risk assessment model; acquiring a current risk control range, wherein the current risk control range is a preset risk control range corresponding to a current resource borrowing type, a resource borrowing duration and resource transaction behavior data; and when the credit score is within the current risk control range, determining that the resource borrowing application data of the current user is approved.

In one embodiment, the processor when executing the computer program further specifically implements the following steps:

the processor executes the computer program to achieve the above-mentioned obtaining of the current risk control range, and the following steps are specifically achieved in the steps:

and retraining the decision tree model according to the new credit score.

training sample data to obtain a rejection sample and a pass sample;

the processor executes the computer program to realize the step of obtaining the audit passing rate of each node in the decision tree model, and specifically realizes the following steps:

receiving a request of a new resource transfer type in a second level;

In one embodiment, the resource lending duration is a duration from a time when the current user applies for the resource corresponding to the current resource lending type to the current time for the first time, and the following steps are specifically implemented when the processor executes the computer program to implement the steps of obtaining the current resource lending type according to the resource lending application data, the resource lending duration corresponding to the current resource lending type, and the resource transaction behavior data between the current user and each resource lending institution within the resource lending duration:

In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of: acquiring resource borrowing application data of a current user; acquiring a current resource lending type, a resource lending duration corresponding to the current resource lending type and resource transaction behavior data between a current user and each resource lending institution within the resource lending duration according to the resource lending application data; obtaining a credit score of the current user according to a preset risk assessment model; acquiring a current risk control range, wherein the current risk control range is a preset risk control range corresponding to a current resource borrowing type, a resource borrowing duration and resource transaction behavior data; and when the credit score is within the current risk control range, determining that the resource borrowing application data of the current user is approved.

In one embodiment, the computer program when executed by the processor further embodies the steps of:

when the computer program is executed by the processor to realize the step of obtaining the current risk control range, the following steps are specifically realized:

and retraining the decision tree model according to the new credit score.

In one embodiment, the computer program, when executed by the processor, further embodies the steps of:

training sample data to obtain a rejection sample and a pass sample;

when the computer program is executed by the processor to realize the step of obtaining the audit passing rate of each node in the decision tree model, the following steps are specifically realized:

receiving a request of a new resource transfer type in a second level;

In one embodiment, the resource lending duration is a duration from a time when the current user applies for the resource corresponding to the current resource lending type to the current time for the first time, and when the computer program is executed by the processor to implement the steps of obtaining the current resource lending type according to the resource lending application data, the resource lending duration corresponding to the current resource lending type, and the resource transaction behavior data between the current user and each resource lending institution within the resource lending duration, the following steps are specifically implemented:

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware related to instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above examples only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. A method of data processing, the method comprising:

acquiring resource borrowing application data of a current user;

acquiring a current resource lending type, resource lending duration corresponding to the current resource lending type and resource transaction behavior data between the current user and each resource lending institution in the resource lending duration according to the resource lending application data;

obtaining the credit score of the current user according to a preset risk assessment model;

acquiring a current risk control range, wherein the current risk control range is a preset risk control range corresponding to the current resource lending type, the resource lending duration and the resource transaction behavior data;

and when the credit score is within the current risk control range, determining that the resource borrowing application data of the current user passes the audit.

2. The method of claim 1, further comprising:

performing model training according to each sample data and credit score of each sample user to obtain a decision tree model, wherein the decision tree model comprises a first level, a second level and a third level, each node in the first level is configured with a corresponding resource borrowing type, each node in the second level is configured with a duration condition corresponding to each resource borrowing type, and each node in the third level is configured with a corresponding risk control range and a resource transaction condition under each duration condition;

the obtaining of the current risk control range includes:

matching the resource lending duration with each duration condition under the current resource lending type, and determining a duration condition matched with the resource lending duration;

and acquiring a risk control range configured on a node corresponding to the resource transaction condition matched with the resource transaction behavior data to obtain the current risk control range.

3. The method of claim 2, further comprising:

comparing the audit passing rate of each node in the decision tree model with the average audit passing rate to obtain a high-passing-rate node and a low-passing-rate node;

retraining the decision tree model based on the new credit score.

4. The method of claim 3, further comprising:

when a high default rate high-pass rate node or a low default rate low-pass rate node exists in the re-trained decision tree model, adjusting risk control ranges corresponding to the high default rate low-pass rate node and the low default rate high-pass rate node according to the default rate and the audit pass rate of the high default rate high-pass rate node, the default rate and the audit pass rate of the low default rate low-pass rate node.

5. The method of claim 3, further comprising:

training the sample data to obtain a rejection sample and a passing sample;

removing a corresponding number of pass samples from the pass samples according to a preset set-out ratio to update the pass samples;

6. The method of claim 2, further comprising:

receiving a request for adding a new resource lending type in the second level;

and extracting a target type in the request of the new added resource borrowing type, and adding the target type into the decision tree model.

7. The method according to claim 1, wherein the resource lending duration is a duration from a time when the current user first applies for the resource corresponding to the current resource lending type to a current time, and the obtaining of the current resource lending type, the resource lending duration corresponding to the current resource lending type, and the resource transaction behavior data between the current user and each resource lending institution within the resource lending duration according to the resource lending application data comprises:

extracting the current resource borrowing type from the resource borrowing application data;

acquiring historical resource lending data of the current user according to the current resource lending type and user information in the resource lending application data;

acquiring the time from the time when the current user applies for the resource corresponding to the current resource transfer type for the first time to the current time according to the historical resource transfer data;

8. A data processing apparatus, characterized in that the apparatus comprises:

a second obtaining module, configured to obtain, according to the resource lending application data, a current resource lending type, a resource lending duration corresponding to the current resource lending type, and resource transaction behavior data between the current user and each resource lending institution within the resource lending duration;

the determining module is used for acquiring a current risk control range, wherein the current risk control range is a preset risk control range corresponding to the current resource lending type, the resource lending duration and the resource transaction behavior data;

9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 7 are implemented when the computer program is executed by the processor.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.