CN114662903A

CN114662903A - Automated credit granting method, apparatus, device, medium, and program product

Info

Publication number: CN114662903A
Application number: CN202210280741.4A
Authority: CN
Inventors: 蔡素贤; 夏成扬; 罗燕龙; 俞泱
Original assignee: China Construction Bank Corp
Current assignee: China Construction Bank Corp
Priority date: 2022-03-21
Filing date: 2022-03-21
Publication date: 2022-06-24

Abstract

The invention relates to an artificial intelligence technology and discloses an automatic credit granting method, a device, equipment, a medium and a program product. The automatic credit granting method comprises the following steps: acquiring financial data, wherein the financial data comprises financial statement data and approval text data of a plurality of samples; inputting the financial statement data and the approval text data into a first classification model to obtain characteristic data contributing to credit approval; and inputting the corresponding characteristic data in the sample into a second classification model according to the characteristic data contributing to the credit approval to obtain a prediction approval decision. Compared with the prior art, the embodiment of the invention enriches the evaluation information quantity of financial indexes and improves the objectivity of credit approval.

Description

Automated credit granting method, apparatus, device, medium, and program product

Technical Field

The present invention relates to the field of artificial intelligence technology, and in particular, to an automated credit granting method, apparatus, device, medium, and program product.

Background

At present, two methods are generally adopted in a trust examination and approval risk monitoring link. One approach is to simply set the rigid requirements for industry or customer admission financial indicators. The method has the problems that only whether the output is abnormal or not can be output to carry out early warning, but the financial related risk in the credit granting process cannot be effectively evaluated due to less information amount, so that whether the approval passes or not cannot be judged in an auxiliary mode. Another method is to build a scoring system of the financial data in a weighting mode in an artificial subjective mode. It is clear that this method is highly subjective. In conclusion, the existing credit granting and approval method has the problems of small evaluation information amount of financial indexes and strong subjectivity.

Disclosure of Invention

The invention provides an automatic credit granting method, device, equipment, medium and program product, which are used for enriching the evaluation information quantity of financial indexes and improving the objectivity of credit granting and approval.

According to an aspect of the invention, an automatic credit granting method is provided, which comprises the following steps:

acquiring financial data, wherein the financial data comprises financial statement data and approval text data of a plurality of samples;

inputting the financial statement data and the approval text data into a first classification model to obtain characteristic data contributing to credit approval;

and inputting the corresponding characteristic data in the sample into a second classification model according to the characteristic data contributing to the credit approval to obtain a prediction approval decision.

Optionally, the method for acquiring the feature data contributing to the approval of the credit specifically includes:

classifying the financial information in the financial statement data to construct various financial characteristics;

analyzing the approval text data and extracting evaluation information to construct a plurality of first black labels and a plurality of first white labels; the first black label or the first white label corresponds to one of the financial features;

and the first classification model calculates the importance ranking of the financial characteristics according to the financial characteristics, the first black label and the first white label to obtain the characteristic data contributing to the credit approval.

Optionally, the financial features comprise: at least one of a characteristic reflecting financial fraud, a characteristic reflecting enterprise financial data quality scores, a characteristic reflecting debt, a characteristic reflecting profit, a characteristic reflecting operation, a characteristic reflecting growth, a characteristic reflecting cash flow capacity and a surplus management index characteristic reflecting measurement profit control level.

Optionally, the method for constructing the financial feature specifically includes:

establishing a scoring rule for the financial statement data, filtering out data with a score lower than a threshold value, and perfecting the rest data;

based on the angle of financial approval business, re-identifying, screening, explaining and analyzing the initial financial characteristics in the financial statement data, and deriving intermediate financial characteristics reflecting the conditions of enterprises;

and merging the initial financial characteristics and the intermediate financial characteristics according to financial functions to obtain the financial characteristics.

Optionally, before establishing the scoring rule, further comprising:

verifying the authenticity, integrity and checking relationship of the financial statement data;

and screening the financial information quality of the financial statement data according to the verification result.

Optionally, the method for refining data includes: at least one of data cleaning, missing value filling, abnormal value processing, feature binning, feature importance checking and feature correlation processing.

Optionally, the first classification model comprises an Xgboost classification model.

Optionally, the second classification model comprises a machine learning classification model.

Optionally, the method for obtaining the machine learning classification model specifically includes:

extracting characteristic data of the plurality of samples contributing to the approval of the credit;

if the sample approval conclusion is a continuation or a rejection, marking the sample as a second black label, and taking the corresponding sample as a black sample; if the sample is approved, marking the sample as a second white label, and taking the corresponding sample as a white sample;

constructing a training set and a testing set according to a rule that the number of the black samples and the number of the white samples are in a first preset proportion;

and training the machine learning classification model according to the training set and the testing set.

Optionally, the method for constructing the training set and the test set specifically includes:

judging whether the number of the black samples and the number of the white samples are balanced or not according to the first preset proportion; if so, extracting black samples and white samples with corresponding quantity from the samples;

otherwise, the black samples are selected in full quantity, and the white samples with corresponding quantity are extracted according to the first preset proportion; constructing a plurality of data sets according to the same rule;

and dividing the data set into a training set and a testing set according to a second preset proportion.

Optionally, the training the machine learning classification model according to the training set and the test set includes:

inputting the training set into a model pool to obtain an output result of the model pool;

inputting the output result of the model pool into a neural network to obtain a machine learning classification model to be evaluated;

and inputting the test set into the machine learning classification model to be evaluated so as to perform model evaluation and subsequent optimization.

Optionally, after obtaining the feature data contributing to the approval of the trust, the method further includes:

and generating and outputting a financial physical examination table, wherein the financial physical examination table comprises the characteristic data type and the numerical range which contribute to the approval of the credit.

According to another aspect of the present invention, there is provided a trust approval apparatus, including:

the data acquisition module is used for acquiring financial data, wherein the financial data comprises financial statement data and approval text data of a plurality of samples;

the first evaluation module is used for inputting the financial statement data and the approval text data into a first classification model to obtain characteristic data contributing to credit approval;

and the second evaluation module is used for inputting the corresponding characteristic data in the sample into a second classification model according to the characteristic data contributing to the credit approval to obtain a prediction approval decision.

According to another aspect of the present invention, there is provided an electronic apparatus including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores a computer program executable by the at least one processor, the computer program being executable by the at least one processor to enable the at least one processor to perform the automated credit granting method according to any of the embodiments of the present invention.

According to another aspect of the present invention, there is provided a computer-readable storage medium storing computer instructions for causing a processor to implement an automated credit granting method according to any of the embodiments of the present invention when executed.

According to another aspect of the invention, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the automated credit granting method according to any of the embodiments of the invention.

According to the embodiment of the invention, financial data is taken as a main body, and by introducing a first classification model and a second classification model, related risk approval judgment of an approval worker is actively learned, and after a better fitting model is trained, the approval result is predicted. The first classification model outputs characteristic data contributing to credit approval, and assessment information quantity of financial indexes can be enriched; and the second classification model is adopted to predict the examination and approval decision, so that the objectivity of receiving and approval can be improved. Therefore, the embodiment of the invention can help the examining and approving personnel to more comprehensively know the financial condition of the enterprise, and the examining and approving result is more objective.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present invention, nor do they necessarily limit the scope of the invention. Other features of the present invention will become apparent from the following description.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

Fig. 1 is a schematic flow chart of an automated credit granting method according to an embodiment of the present invention;

FIG. 2 is a flow chart illustrating another method for automated credit reporting according to an embodiment of the invention;

FIG. 3 is a schematic diagram illustrating a first phase according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of a method of constructing a financial feature according to an embodiment of the present invention;

FIG. 5 is a diagram illustrating a second stage of executing steps according to an embodiment of the present invention;

FIG. 6 is a diagram illustrating training of a machine learning classification model according to an embodiment of the present invention;

fig. 7 is a schematic structural diagram of a trust approval apparatus according to an embodiment of the present invention;

fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in other sequences than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

It should be noted that in the embodiments of the present invention, the acquisition, storage and/or processing of the financial data involved complies with relevant regulations of national laws and regulations, and does not violate the customs of the public order.

The embodiment of the invention provides an automatic credit granting method which is suitable for a link of risk monitoring of approval of credit granting and credit granting, and assists relevant personnel in effectively evaluating financial related risks in a credit granting process by constructing classification characteristics and adopting a first classification model and a second classification model. The method can be executed by a credit approval device, the device can be integrated in electronic equipment such as a computer, a mobile phone, a tablet computer and the like, and the device can be executed by software and/or hardware.

Fig. 1 is a schematic flow chart of an automated credit granting method according to an embodiment of the present invention. Referring to fig. 1, the automated credit granting method includes the steps of:

and S110, acquiring financial data, wherein the financial data comprises financial statement data and approval text data of a plurality of samples.

The obtained financial data can be original financial data or processed financial data. If the original financial data is acquired, the financial data needs to be preprocessed. Illustratively, the raw financial data includes: the system comprises data of approval opinions and approval conclusion data approved by credit line approval, financial statement data of an enterprise approving business in three years, industry information of the enterprise, industry and commerce information of the enterprise and the like. The rules for data preprocessing may include: the currency is RMB; the financial and newspaper enterprise type is enterprise and non-group; the financial report type is annual report; the financial and newspaper deadline date range is 20XX-XX-XX to 20YY-YY-YY, and the latest financial and newspaper deadline date is obtained by considering the update state and the audit condition according to the client number and the financial and newspaper deadline date; taking out all financial statement fields from the original financial data, calculating the loss rate of the financial statement fields, and deleting the fields if the loss rate exceeds 80%; in order to avoid the influence of an extreme value on the model, tail reduction processing of 1% percentile and 99% percentile is carried out on each field, namely after 1% quantiles and 99% quantiles of the field are calculated, the quantiles smaller than 1% are replaced by the quantiles 1%, and the quantiles larger than 99% are replaced by the quantiles 99%.

And S120, inputting the financial statement data and the approval text data into the first classification model to obtain characteristic data contributing to credit approval.

Wherein, financial statement data can be used to form financial characteristics, and the approval text data can be used to mark the financial characteristics as a first black label or a first white label. Optionally, the first classification model comprises an Xgboost classification model. The Xgboost is called eXtrememe Gradient Boosting, and an optimized distributed Gradient Boosting library aims to be efficient, flexible and portable. Xgboost solves both the classification and regression problems. Illustratively, when constructing the Xgboost classification model, the ratio of black and white samples may be 1: 10 the data is extracted. Training set 20XX-20YY years, testing set 20ZZ years.

Specifically, there are three ways of calculating feature importance for Xgboost: weight, gain and cover. weight, the sum of the number of nodes for dividing the characteristics in all the trees; gain, the average gain characterizing a segmentation; it can be understood as the sum of the second derivatives of the samples that are assigned to the node, and the criterion of the feature metric is the average coverage value. Illustratively, the importance ranking of the features is output by analyzing the importance rankings of the features calculated by three methods and selecting a proper method by combining business related knowledge.

And S130, inputting the corresponding characteristic data in the sample into a second classification model according to the characteristic data contributing to the credit approval to obtain a prediction approval decision.

The feature data contributing to the approval of the credit is the features extracted by the first classification module, for example, the financial features of the TOP30 of contribution degree (TOP 30). And each sample can be marked as a second black label or a second white label so as to distinguish the black and white samples, thereby training a second classification model. Optionally, the second classification model comprises a machine learning classification model. And inputting the financial data to be approved into the second classification model to obtain a prediction approval decision for reference of related approving personnel.

In summary, in the embodiment of the present invention, financial data is used as a main body, and by introducing the first classification model and the second classification model, the related risk approval and judgment of the approval staff is actively learned, and after a better fitting model is trained, the approval result is predicted. The first classification model outputs characteristic data contributing to credit approval, and assessment information quantity of financial indexes can be enriched; and the second classification model is adopted to predict the examination and approval decision, so that the objectivity of receiving and approval can be improved. Therefore, the embodiment of the invention can help the examining and approving personnel to more comprehensively know the financial condition of the enterprise, and the examining and approving result is more objective.

Fig. 2 is a schematic flow chart of another automated credit granting method according to an embodiment of the present invention. Referring to fig. 2, on the basis of the foregoing embodiments, optionally, after S120, the method further includes:

and S121, generating and outputting a financial physical examination table, wherein the financial physical examination table comprises the characteristic data types and the numerical value ranges contributing to the approval of the credit.

The financial physical examination table is output, namely, an application program is constructed and output to display equipment such as a display and a printer, and the financial physical examination table is visualized, so that an approver can see the type and the numerical range of the characteristic data which is output by the first classification model and contributes to approval and approval of the approval. The setting can prompt the examining and approving personnel to pay key attention to the related characteristic data, so that the examining and approving personnel can be helped to make decisions.

It can be seen from the above embodiments that the automated credit granting method provided by the embodiments of the present invention includes two stages, where the first stage obtains feature data contributing to credit approval according to the first classification model, and outputs a financial physical examination table. And in the second stage, according to the characteristic data which is extracted in the first stage and contributes to the approval of the credit, a second classification model is constructed to predict whether the subsequent approval passes or not. The execution steps of the first stage and the second stage are specifically described below.

Fig. 3 is a schematic diagram illustrating a first stage of executing steps according to an embodiment of the present invention. Referring to fig. 3, in an embodiment of the present invention, optionally, the step of performing the first stage includes:

and S210, acquiring original financial data.

Wherein this step may be combined with S110 in the foregoing embodiment.

And S220, preprocessing data.

This step may be combined with S110 in the foregoing embodiment, or may be performed separately. If the data preprocessing has been performed in S110, the step is combined with S110; if the data preprocessing is not performed in S110, the step is performed separately. Illustratively, the rules of data pre-processing may include: the currency is RMB; the financial and newspaper enterprise type is enterprise and non-group; the financial report type is annual report; the financial and newspaper deadline date range is 20XX-XX-XX to 20YY-YY-YY, and the latest financial and newspaper deadline date is obtained by considering the updating state and the auditing condition according to the client number and the financial and newspaper deadline date; taking out all financial statement fields from the original financial data, calculating the loss rate of the financial statement fields, and deleting the fields if the loss rate exceeds 80%; in order to avoid the influence of an extreme value on the model, tail reduction processing of 1% percentile and 99% percentile is carried out on each field, namely after 1% quantiles and 99% quantiles of the field are calculated, the quantiles smaller than 1% are replaced by the quantiles 1%, and the quantiles larger than 99% are replaced by the quantiles 99%.

And S230, classifying the financial information in the financial statement data, and constructing various financial characteristics.

The financial statement data comprises different types of financial information, but the classification of the financial information is often not in accordance with the approval requirement of credit granting. Therefore, there is a need to merge existing features according to their function, and to make sure that each feature reflects what aspect of the enterprise's capabilities or behavior. Exemplary financial characteristics include: at least one of a characteristic reflecting financial fraud, a characteristic reflecting enterprise financial data quality scores, a characteristic reflecting debt, a characteristic reflecting profit, a characteristic reflecting operation, a characteristic reflecting growth, a characteristic reflecting cash flow capacity and a surplus management index characteristic reflecting measurement profit control level.

S240, obtaining the credit approval and approval opinions.

The approval reply opinions are contents in approval text data and belong to the text data.

And S250, analyzing and extracting evaluation information.

Since the credit approval review opinions belong to the text data, the text data needs to be analyzed and evaluation information needs to be extracted to construct a subsequent black-and-white label.

And S260, constructing a black-and-white label.

The method comprises the steps that a black-and-white label is constructed according to extracted evaluation information, the extracted evaluation information comprises evaluation information with different characteristics, and for example, financial characteristics of the evaluation information, which reflect overhigh liability and liability of assets, poor profit capacity of major business and the like, are marked as a first black label; the financial characteristics without evaluation information such as too high rate of assets and liabilities, poor profit capacity of main business and the like are marked as a first white label. I.e. the first black label or the first white label corresponds to a financial feature.

And S270, inputting the financial characteristics, the first black label and the first white label into the first classification model.

Wherein the first classification model is an Xgboost classification model. The Xgboost classification model is an optimized distributed gradient promotion library, and aims to be efficient, flexible and portable. Xgboost solves both the classification and regression problems. Specifically, there are three ways of calculating feature importance for Xgboost: weight, gain and cover. weight, the sum of the number of nodes for dividing the characteristics in all the trees; gain, the average gain characterizing a segmentation; it can be understood as the sum of the second derivatives of the samples that are assigned to the node, and the criterion of the feature metric is the average coverage value. Illustratively, the importance ranking of the feature is calculated by analyzing three methods, and an appropriate method is selected by combining business related knowledge to output the importance ranking of the feature.

And S280, outputting the importance ranking of the features.

Illustratively, the financial characteristics of the degree of contribution TOP30 are output.

And S290, outputting a financial physical examination table.

Illustratively, based on the financial characteristics of the contribution degree TOP30 output by S280, a reasonable interval of each financial characteristic is calculated, and a financial physical examination table of the enterprise is formed. The larger the financial index is, the better the financial index is, the forward index is called; the smaller the value, the better the value, which is called negative indicator. The range value method of the reasonable interval is as follows: from the white samples, the 99% quantile (d1), 1% quantile (d2), mean +3 times standard deviation value (d3), mean-3 times standard deviation value (d4) of each index were calculated, the forward index: the upper bound is min (d1, d3) and the lower bound is max (d2, d 4). Negative direction index: the upper bound is min (d2, d4) and the lower bound is max (d1, d 3).

The execution steps of the first stage are realized by S210-S290. In the first stage, the financial characteristics concerned by the examining and approving personnel are extracted mainly by constructing black and white labels of the financial characteristics and financial problems; and constructing an application based on the characteristics to assist in the financial condition analysis and output a financial physical examination table. By the arrangement, for enterprises needing to pass credit approval, financial allocation can be reasonably carried out by paying attention to the financial physical examination table; for the approvers who need to examine and approve the credit approval, the reference of the approval can be carried out by paying attention to the financial physical examination table.

On the basis of the foregoing embodiments, optionally, the method for constructing the financial characteristics specifically includes: and establishing a scoring rule for the financial statement data, filtering out data with a score lower than a threshold value, and perfecting the rest data. Based on the financial approval business, the initial financial characteristics in the financial statement data are re-identified, screened, explained and analyzed, and intermediate financial characteristics reflecting the conditions of the enterprise are derived. And merging the initial financial characteristics and the intermediate financial characteristics according to the financial function to obtain the financial characteristics. Specifically, fig. 4 is a schematic diagram of a method for constructing a financial feature according to an embodiment of the present invention. Referring to fig. 4, the method of constructing the financial features specifically includes:

and S211, verifying data.

The data verification comprises the verification of authenticity, integrity and checking relation of the financial statement data.

And S212, scoring the quality of the enterprise financial data.

And establishing a scoring rule for the financial statement data according to the verification result so as to score the quality of the enterprise financial data and discriminate the financial information quality of the financial statement data.

And S213, completing the data.

And refining the data refers to filtering out data with a score lower than a threshold value and refining the rest data. Illustratively, ways to refine the data include: at least one of data cleaning, missing value filling, abnormal value processing, feature binning, feature importance checking and feature correlation processing.

S214, feature derivation and service interpretation.

The characteristic derivation and the business explanation refer to re-identifying, screening, explaining and analyzing initial financial characteristics in the financial statement data based on the angle of financial approval business, and deriving intermediate financial characteristics reflecting the conditions of the enterprise.

S215, merging the characteristics.

The characteristic merging means that the initial financial characteristic and the intermediate financial characteristic are merged according to financial functions to obtain the financial characteristic. These financial characteristics should be those that enable enterprise capability system settings. In particular, the financial features after feature merging can make clear which aspect of the enterprise's capabilities or behaviors each feature reflects. The system mainly comprises field characteristics reflecting enterprise repayment, profit, operation, cash flow capacity and the like, and surplus management indexes reflecting enterprise profit control levels (because information in two aspects needs to be considered in the crediting and approval process, the financial data quality of an enterprise to be approved is important, and the degree of the enterprise to subjectively regulate and control the profit represents the integrity of financial information and the objective compliance of the financial information represents the later under a certain level of the financial data quality). The features are divided into surplus management indexes reflecting financial fraud, enterprise financial data quality scores, debt repayment, profit, operation, growth, cash flow capacity and metering profit control levels, and the like, and the surplus management indexes have 8 dimensions in total.

The financial characteristics are constructed through S211-S215, the financial characteristics constructed through analyzing and extracting the financial data can clearly show the capability or behavior of which aspect of the enterprise each characteristic reflects, the automation degree is high, manual classification by financial staff is not needed, and the method has the characteristics of high accuracy and high efficiency.

Fig. 5 is a schematic diagram illustrating a second stage of executing steps according to an embodiment of the present invention. Referring to fig. 5, in an embodiment of the present invention, optionally, the method for obtaining the machine learning classification model specifically includes:

and S310, extracting characteristic data of the plurality of samples, wherein the characteristic data contributes to credit approval.

Illustratively, the financial features of the contribution TOP30, which are the features acquired in the first stage, are extracted.

And S320, marking the sample.

If the sample approval conclusion is a continuation or a rejection, marking the sample as a second black label, and taking the corresponding sample as a black sample; and marking the sample as a second white label if the sample passes the approval conclusion, wherein the corresponding sample is a white sample. Unlike the first black and white label of the first stage, which is for a financial characteristic, the second black and white label is for a sample.

And S330, inputting a machine learning classification model.

And constructing a training set and a testing set according to a rule that the number of the black samples and the number of the white samples are in a first preset proportion, and training a machine learning classification model according to the training set and the testing set.

And S340, outputting the examination and approval result prediction.

And in the process of carrying out the machine learning classification model, the input quantity is financial characteristics and sample black and white labels, and the output quantity is an approval result. According to the existing training set, the machine learning classification model can be constructed. After the machine learning classification model is built, the financial data to be approved is input into the machine learning classification model, and approval result prediction of the financial data to be approved can be obtained.

Through S310-S340, the construction of the machine learning classification model is completed.

Fig. 6 is a schematic diagram of training a machine learning classification model according to an embodiment of the present invention. Referring to fig. 6, on the basis of the foregoing embodiments, optionally, the method for constructing the training set and the test set specifically includes:

and S331, constructing a data set.

The construction of the data set judges whether the number of the black samples and the number of the white samples are balanced or not according to a first preset proportion; if so, extracting black samples and white samples with corresponding quantity from the samples; otherwise, selecting the black samples as the total amount, and extracting a corresponding amount of white samples according to a first preset proportion; multiple data sets are constructed with the same rules. Illustratively, the first predetermined ratio is 1, 3, 5, 10, etc. Taking the case that black and white samples are unbalanced, 5 data sets are constructed as an example, black samples in the data set 1 are selected as full quantities, and white samples are sampled according to 5 times of the black samples; selecting black samples in the data set 2 as a full amount, and sampling white samples according to 5 times of the black samples; … …; the black samples in data set 5 were selected for the full volume and the white samples were sampled at 5 times the black samples.

S332, constructing a training set and a testing set.

The training set and the test set are obtained by taking the data set as a basis, and the data set is divided into the training set and the test set according to a second preset proportion. Illustratively, the second preset ratio is 7: 3 in equal proportion. Specifically, 70% of the data set 1 is the training set and 30% is the testing set; 70% of the data set 2 is the training set and 30% is the testing set; … …, respectively; 70% of the data set 5 is the training set and 30% is the test set.

Training a machine learning classification model according to a training set and a testing set, comprising:

and S341, inputting a model pool.

Specifically, the training set is input into a model pool to obtain an output result of the model pool. Illustratively, the model pool may include: logistic regression model, random forest model, GBDT model, Xgboost model, etc. If the model pool comprises the four models, 20(5 x 4) prediction results are trained, then the interior of the models is unified, and the prediction results are averaged.

And S342, inputting the neural network.

Specifically, the output result of the model pool is input into a neural network, and a machine learning classification model to be evaluated is obtained. If the model pool comprises the four models, the result is input into a neural network by taking the mean value, and four model weights are calculated by each type of model (a logistic regression model, a random forest model, a GBDT model and an Xgboost model).

And S343, evaluating the model.

Specifically, the test set is input into a machine learning classification model to be evaluated so as to perform model evaluation and subsequent tuning. Illustratively, the composite model is tested with test set data and evaluated, and subsequently tuned, primarily with respect to Precision (Precision) and Recall (Recall) values, and AUC values.

The construction of the machine learning classification model is completed through S331-S343.

In summary, in the embodiment of the present invention, financial data is used as a main body, and by introducing the first classification model and the second classification model, the examining and approving personnel is actively learned to examine and determine the related risks, and after a better fitting model is trained, the examining and approving result is predicted. The first classification model outputs characteristic data contributing to credit approval, and assessment information quantity of financial indexes can be enriched; and the second classification model is adopted to predict the examination and approval decision, so that the objectivity of receiving and approval can be improved. Therefore, the embodiment of the invention can help the examining and approving personnel to more comprehensively know the financial condition of the enterprise, and the examining and approving result is more objective.

The embodiment of the invention also provides a credit approval device. Fig. 7 is a schematic structural diagram of a trust approval apparatus according to an embodiment of the present invention. Referring to fig. 7, the approval apparatus includes:

the data acquisition module 410 is used for acquiring financial data, wherein the financial data comprises financial statement data and approval text data of a plurality of samples;

the first evaluation module 420 is used for inputting the financial statement data and the approval text data into the first classification model to obtain characteristic data contributing to the credit approval;

and the second evaluation module 430 is configured to input the corresponding feature data in the sample into the second classification model according to the feature data contributing to the approval of the credit, so as to obtain a prediction approval decision.

Optionally, the first evaluation module specifically includes:

the characteristic construction unit is used for classifying the financial information in the financial statement data and constructing various financial characteristics;

the first marking unit is used for analyzing the approval text data and extracting evaluation information to construct a plurality of first black labels and a plurality of first white labels; the first black label or the first white label corresponds to a financial characteristic;

and the contribution characteristic screening unit is used for calculating the importance ranking of the financial characteristics according to the financial characteristics, the first black label and the first white label by the first classification model to obtain characteristic data which contributes to the credit approval.

Optionally, the feature construction unit is specifically configured to:

based on the angle of financial approval business, re-identifying, screening, explaining and analyzing the initial financial characteristics in the financial statement data, and deriving intermediate financial characteristics reflecting the conditions of the enterprise;

and merging the initial financial characteristics and the intermediate financial characteristics according to the financial function to obtain the financial characteristics.

The characteristic construction unit is also used for verifying the authenticity, integrity and checking relationship of the financial statement data before establishing the grading rule;

Optionally, the method for refining data includes: data cleaning, missing value filling, abnormal value processing, characteristic binning, characteristic importance testing and characteristic correlation processing.

Optionally, the obtaining module of the machine learning classification model specifically includes:

extracting characteristic data of a plurality of samples contributing to credit approval;

the marking unit is used for marking the sample as a second black label if the approval conclusion of the sample is a continuation or a rejection, and the corresponding sample is a black sample; if the sample is approved, marking the sample as a second white label, and taking the corresponding sample as a white sample;

the training set and test set constructing unit is used for constructing a training set and a test set according to a rule that the number of black samples and the number of white samples are in a first preset proportion;

and the training unit is used for training the machine learning classification model according to the training set and the test set.

Optionally, the training set and test set constructing unit is specifically configured to:

judging whether the number of the black samples and the number of the white samples are balanced or not according to a first preset proportion; if so, extracting black samples and white samples with corresponding quantity from the samples;

otherwise, selecting the black samples as the total amount, and extracting a corresponding amount of white samples according to a first preset proportion; constructing a plurality of data sets by the same rule;

Optionally, the training unit is specifically configured to:

and inputting the test set into a machine learning classification model to be evaluated so as to evaluate the model and perform subsequent optimization.

Optionally, the first evaluation module is further configured to generate and output a financial physical examination table after obtaining the feature data contributing to the approval of the credit, where the financial physical examination table includes the type and the value range of the feature data contributing to the approval of the credit.

The trust approval device provided by the embodiment of the invention can execute the automatic trust method provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

Fig. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed herein.

As shown in fig. 8, the electronic device 10 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a Read Only Memory (ROM)12, a Random Access Memory (RAM)13, and the like, wherein the memory stores a computer program executable by the at least one processor, and the processor 11 can perform various suitable actions and processes according to the computer program stored in the Read Only Memory (ROM)12 or the computer program loaded from a storage unit 18 into the Random Access Memory (RAM) 13. In the RAM 13, various programs and data necessary for the operation of the electronic apparatus 10 may also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. An input/output (I/O) interface 15 is also connected to bus 14.

A number of components in the electronic device 10 are connected to the I/O interface 15, including: an input unit 16 such as a keyboard, a mouse, or the like; an output unit 17 such as various types of displays, speakers, and the like; a storage unit 18 such as a magnetic disk, an optical disk, or the like; and a communication unit 19 such as a network card, modem, wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.

The processor 11 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various processors running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, or the like. Processor 11 performs the various methods and processes described above, such as an automated credit granting method.

In some embodiments, the automated crediting method may be implemented as a computer program product that is tangibly embodied in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and/or installed onto the electronic device 10 via the ROM 12 and/or the communication unit 19. When loaded into RAM 13 and executed by processor 11, the computer program may perform one or more of the steps of the automated crediting method described above. Alternatively, in other embodiments, the processor 11 may be configured to perform the automated crediting method by any other suitable means (e.g., by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

A computer program for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions/acts specified in the flowchart and/or block diagram block or blocks to be performed. A computer program can execute entirely on a machine, partly on a machine, as a stand-alone software package partly on a machine and partly on a remote machine or entirely on a remote machine or server.

In the context of the present invention, a computer-readable storage medium may be a tangible medium that can contain, or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. A computer readable storage medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. Alternatively, the computer readable storage medium may be a machine readable signal medium. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the electronic device. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), blockchain networks, and the internet.

The computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical host and VPS service are overcome.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present invention may be executed in parallel, sequentially, or in different orders, and are not limited herein as long as the desired results of the technical solution of the present invention can be achieved.

The above-described embodiments should not be construed as limiting the scope of the invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An automated credit granting method, comprising:

2. The method according to claim 1, wherein the method for obtaining the feature data contributing to the approval of the credit specifically comprises:

3. The method of claim 2, wherein the financial characteristics comprise: at least one of a characteristic reflecting financial fraud, a characteristic reflecting enterprise financial data quality scores, a characteristic reflecting debt, a characteristic reflecting profit, a characteristic reflecting operation, a characteristic reflecting growth, a characteristic reflecting cash flow capacity and a surplus management index characteristic reflecting measurement profit control level.

4. The method according to claim 2, characterized in that the method of construction of the financial features comprises in particular:

based on the angle of financial examination and approval business, re-identifying, screening, explaining and analyzing the initial financial characteristics in the financial statement data, and deriving intermediate financial characteristics reflecting the conditions of enterprises;

5. The method of claim 4, further comprising, prior to establishing the scoring rules:

6. The method of claim 4, wherein refining the data comprises: at least one of data cleaning, missing value filling, abnormal value processing, feature binning, feature importance checking and feature correlation processing.

7. The method of any of claims 1-6, wherein the first classification model comprises an Xgboost classification model.

8. The method of any of claims 1-6, wherein the second classification model comprises a machine learning classification model.

9. The method according to claim 8, wherein the method for obtaining the machine learning classification model specifically comprises:

10. The method according to claim 9, wherein the method for constructing the training set and the test set specifically comprises:

11. The method of claim 9, wherein training the machine learning classification model according to the training set and the test set comprises:

12. The method of claim 1, further comprising, after obtaining the feature data contributing to the approval of the trust:

13. A credit approval device, comprising:

14. An electronic device, characterized in that the electronic device comprises:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein, the first and the second end of the pipe are connected with each other,

the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the automated credit granting method of any one of claims 1-12.

15. A computer-readable storage medium storing computer instructions for causing a processor to perform the automated credit granting method of any one of claims 1-12 when executed.

16. A computer program product, characterized in that the computer program product comprises a computer program which, when being executed by a processor, implements an automated credit granting method according to any one of claims 1-12.