CN110400208B - Small and micro risk control model construction method and application method - Google Patents
Small and micro risk control model construction method and application method Download PDFInfo
- Publication number
- CN110400208B CN110400208B CN201810379596.9A CN201810379596A CN110400208B CN 110400208 B CN110400208 B CN 110400208B CN 201810379596 A CN201810379596 A CN 201810379596A CN 110400208 B CN110400208 B CN 110400208B
- Authority
- CN
- China
- Prior art keywords
- data
- portrait
- data group
- forming
- risk control
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Physics & Mathematics (AREA)
- Finance (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Accounting & Taxation (AREA)
- Educational Administration (AREA)
- Technology Law (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention relates to the technical field of financial risk control, in particular to a construction method and an application method of a small and micro risk control model, wherein the construction method of the small and micro risk control model comprises the steps of screening and forming a sample database in an initial client image library according to a preset method, and forming a determined data group according to the sample database; calculating to form derived data group data according to the determined data group based on EM-tool algorithm; and forming deep learning group data according to the derived data group and the determined data group based on a vector machine algorithm.
Description
Technical Field
The invention relates to the technical field of financial risk control, in particular to a construction method and an application method of a small and micro risk control model.
Background
The starting point of the Chinese commercial Min-finance is in 2005, which is originated from the world bank initiative, the national development bank undertaking, and the German IPC company providing technical support [ the national development bank Min-Credit project ]. The project introduces the commercial small-micro credit wind control technology into China for the first time, and under the project, 18 city commercial banks and rural commercial banks are supported to carry out small-micro credit business and propagate IPC small-micro credit technology, such as a Taoist counter-state bank, a contracting bank, a Chongqing bank, a Guiyang bank, an Anhui Maanshan business and the like in the field. Since 2008, a large number of city commercial banks, rural credit unions and small credit companies introduce and propagate IPC small credit wind control technology, which becomes the core wind control technology for various financial institutions to develop small credit business.
The core of the existing IPC (International patent computer) minor credit wind control technology is to restore financial statements of minor merchants without statement informality through a cross check technology and carry out rapid loan approval through widely authorized manual loan bank examination meetings. The IPC small and micro wind control technology can accurately calculate the first repayment capacity of the client to avoid long debt and evaluate the repayment willingness of the client, so that the credit risk is controlled before credit. Taking the practice of the IPC credit technology practiced by the Taizhou bank as an example, the small micro business of the Taizhou bank keeps below 1% of wind control level for a long time for more than 10 years, and the effectiveness of the IPC small micro credit technology is demonstrated by using high-quality asset quality.
The IPC mini-Credit wind control technology has some obvious disadvantages in the process of popularization for many years, such as: firstly, a 'full-function' client manager is seriously relied on, the 'full-function' client manager usually needs a longer training period, the training entry time of the client manager is usually 3 months, the basic mastery is 6 months, and 12 months are needed for fully developing various services, so that an organization wishing to quickly form the service scale needs a very long technical reserve period. Meanwhile, when a client manager can independently lead a team, the client manager inherits the corresponding auditing technology through the formation of 'the master carries the higer', but one master is difficult to comprehensively master all the auditing technologies or skills, so that the inherited technologies or skills are continuously shrunk in the later-stage technical inheritance, and a vicious circle is formed repeatedly and continuously; and secondly, the IPC small and micro wind control technology takes a customer manager as a center, the operation risk in the credit process is higher, and the customer manager finishes marketing, investigation, tabulation and post-loan management by one person. The threshold of a slightly experienced customer manager and the traditional data and data counterfeiting of the customer is low, and how to prevent the moral risk prevention of the small customer manager is often the pain point of each financial institution applying the IPC technology; and thirdly, the on-line loan approval system of the IPC small and micro wind control technology highly depends on letter examiners with abundant wind control experience, and the approval efficiency is low. Generally, the examination and approval authority of the IPC offline examination and approval conference is strictly determined according to the respective authority combination of the examiners, so that the participation of personnel with higher examination and approval authority in the branch office is often caused, or the examination and approval personnel with investigation authority cannot examine and approve the loans with the association due to the principle of separation of examination and approval, and the examination and approval inefficiency is caused.
Disclosure of Invention
In view of the above problems, embodiments of the present invention provide a method for constructing a small and micro risk control model and an application method thereof, which aim to improve auditing efficiency, reduce risks and reduce labor costs.
In one aspect, the present disclosure provides a method for constructing a small breeze risk control model, including:
screening and forming a sample database in the initial customer image library according to a preset method, and forming a determined data group according to the sample database;
calculating to form derived data group data according to the determined data group based on EM-tool algorithm;
and forming deep learning group data according to the derived data group and the determined data group based on a vector machine algorithm.
Preferably, the method for constructing a small risk control model includes: screening and forming a sample database in the initial client image library according to a predetermined method, and forming a determined data group according to the sample database, wherein the step of forming the determined data group comprises the following steps:
s11, setting M first-class index items and N second-class index items, wherein the first-class index items and the second-class index items are configured with Q option values; forming an initial customer portrait library and portrait combination according to the option values of the first type of index items and the option values of the second type of index items;
step S12, reading any customer portrait data in the initial customer portrait library, and analyzing and judging the current customer portrait data to form a judgment result;
step S13, forming and updating the initial image library according to the judgment result, reading the image combination quantity rejected according to the judgment result, and executing step S12 in a state that the rejected image combination quantity is not less than a preset value; otherwise, forming the sample database according to the updated initial image database.
Preferably, the method for constructing a small risk control model includes: in step S12, reading any client image data in the initial client image library, and analyzing and judging the current client image data to form a judgment result specifically includes:
step S121, in the state that the current client portrait data is judged to be rejected, a combination pool lower than the current client portrait data is eliminated;
step S122, when the client image data is judged to be approved, accepting a combination pool higher than the current client image data.
Preferably, the above method for constructing a small micro-risk control model, wherein the step of calculating and forming derived data group data according to the determined data group based on the EM-tool algorithm specifically includes:
reading a passing rate matched with each option value of the combination line of each portrait data in the determined data set, and acquiring the passing rate of the current portrait data combination line based on the passing rate of each option value;
and sequencing the passing rate of each portrait combination line to form the data of the derived data group.
Preferably, the above method for constructing a small risk control model, based on a vector machine algorithm, forming a deep learning group data according to the derived data group and the determined data group includes:
reading the determined data group data and the derived data group data, and performing weight processing on the derived data group data to form weight derived data group data;
and performing vector machine processing on the determined data group and the weight derived data group to form the deep learning group data.
On the other hand, the present invention provides an application method of a mini-risk control model, wherein the risk control model formed by the mini-risk control model construction method according to any one of the above descriptions further includes:
reading the portrait data to be evaluated of a user, and calculating to obtain the highest credit limit according to the portrait data to be evaluated and the risk control model;
approving the loan request of the current user under the condition that the highest loan limit is not less than the target loan limit of the user;
and forming a suggested loan amount according to the highest lendable amount under the condition that the highest lendable amount is smaller than the target loan amount of the user.
Compared with the prior art, the invention has the advantages that:
the modeling idea of constructing an expert model through subjectively setting index items and manually distributing weights is abandoned, in order to greatly provide the accuracy of the expert model, an initial client portrait library is constructed or updated firstly, and experts participating in modeling are made to perform 0 or 1 judgment based on the initial client portrait library to form a determined data group. Secondly, 70% of derived data group data is obtained through an EM-tool algorithm based on the determined data group, and finally the remaining 30% of calculation is completed through deep learning of the determined data group data and the derived data group data, so that on one hand, subjectivity of expert judgment is reduced to the lowest, and therefore modeling precision is greatly improved, on the other hand, a large number of experiential creditors are not required to carry out approval, approval efficiency is greatly improved, and meanwhile, approval cost is reduced.
Drawings
FIG. 1 is a flowchart of a method for constructing a small risk control model according to an embodiment of the present invention;
FIG. 2 is a flowchart of a method for constructing a small risk control model according to an embodiment of the present invention;
FIG. 3 is a flowchart of a method for constructing a small risk control model according to an embodiment of the present invention;
FIG. 4 is a flowchart of a method for constructing a small risk control model according to an embodiment of the present invention;
FIG. 5 is a flowchart illustrating a method for constructing a small risk control model according to an embodiment of the present invention;
fig. 6 is a flowchart of an application method of the small risk control model according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be further noted that, for the convenience of description, only some of the structures related to the present invention are shown in the drawings, not all of the structures.
Example one
Fig. 1 is a flowchart of a method for constructing a small risk control model according to a first embodiment of the present invention, which can be applied to any mobile terminal. The method may be performed by a server, which may be implemented in software and/or hardware. As shown in fig. 1, a flowchart of a method for constructing a small risk control model according to a first embodiment of the present invention is provided, where the method specifically includes:
s1, screening and forming a sample database in the initial client image library according to a preset method, and forming a determined data group according to the sample database; the method specifically comprises the following steps:
as shown in fig. 2, in step S11, M first-type index items and N second-type index items are set, where the first-type index items and the second-type index items are configured with Q option values; forming an initial customer portrait library and portrait combination according to the option values of the first type of index items and the option values of the second type of index items; the first type of index items can be qualitative index items, such as marital conditions, living conditions and the like, the second type of index items can be quantitative index items, such as liability rate, monthly disposable income data, loan installment repayment amount and the like, the sum of M and N is at least 10, wherein M is a natural number, and N is a natural number; q option values are configured in the M first-type index items or the N second-type index items, the value range of Q can be 2-4, the first-type index items and the second-type index items in each customer portrait data correspond to one option value, namely, each customer portrait data is formed by the option values of the first-type index items and the second-type index items.
Step S12, reading any customer portrait data in the initial customer portrait library, and analyzing and judging the current customer portrait data to form a judgment result; each customer portrait data includes M results of the first type index items and N option values of the second type index items, and the determination result may be completed by an expert or a calculator, which is not limited herein. As shown in fig. 3, the method specifically includes:
step S121, in the state that the current client portrait data is judged to be rejected, a combination pool lower than the current client portrait data is eliminated; the passing probability of all the client portrait data in the combined pool lower than the current client portrait data is smaller than that of the current client portrait data, so that the client portrait data is directly eliminated.
Step S122, when the client image data is judged to be approved, accepting a combination pool higher than the current client image data. All client portrait data in the combined pool higher than the current client portrait data have no passing probability smaller than that of the current client portrait data, so that the client portrait data can be directly accepted. After step S121 or step S122, a client image combination line is obtained, and the number of client images in the initial client image library can be further reduced according to the client image combination line.
Step S13, forming and updating the initial image library according to the judgment result, reading the image combination quantity rejected according to the judgment result, and executing step S12 in a state that the rejected image combination quantity is not less than a preset value; otherwise, forming the sample database according to the updated initial image database. Wherein the predetermined number is 5-10, that is, the judgment is stopped in the state that only 5-10 image combinations can be eliminated after each judgment. After each judgment, under the state that the number of the portrait combinations discarded is more than 10, it can be judged that a large amount of redundant data still exists in the current initial portrait library, and at this time, further discarding processing needs to be performed according to the current customer portrait combination line. Stopping the determination when only 5-10 image combinations can be removed after each determination, wherein part of data to be removed still exists in the initial image library, but the removal efficiency is relatively low, so the determination is stopped.
Step S2, calculating and forming derived data group data according to the determined data group based on EM-tool algorithm; the method specifically comprises the following steps: as shown in fig. 4, step S21 reads the passing rate of each option value in the determination data set matching with the combined line of each image data, and obtains the passing rate of the combined line of the image data based on the passing rate of each option value;
step S22, sorting the passing rate of each of the portrait combination lines to form the derived data set.
And step S3, forming deep learning group data according to the derived data group and the determined data group based on a vector machine algorithm. Specifically, as shown in fig. 5, the method includes:
step S31, reading the determined data group and the derived data group, and performing weighting processing on the derived data group to form a weighted derived data group; the weight of the derived data group data can be 0.2-1.5.
And step S32, performing vector machine processing according to the determined data group and the weight derived data group to form the deep learning group data.
The machine may also perform the judgment by taking a manual judgment to form a judgment result, which is similar to the manual judgment principle, and only the manual judgment is taken as an example here, and the corresponding expert choice is first screened out.
In the construction of the small and micro risk control model, 6 credit and audit experts participate in manual judgment and modeling, and each credit and audit expert has at least 5 years of offline audit experience. Firstly, determining the weight of each expert, determining the final expert weight by adopting 3 indexes, and calculating the values of three indexes, namely identifiability F1 of a data mode corresponding to each expert marking data, stability F2 of the data and expert service comprehensive capacity F3, wherein F1 and F2 are objective indexes and can be obtained by directly calculating through the data; f3 is a subjective index, the evaluation is carried out by adopting a manual evaluation method, a plurality of reviewers are selected, each reviewer knows about each expert, the comprehensive expert service capability corresponding to each expert is obtained through the review of a plurality of experts, the importance of the three indexes is the same, and the three indexes are reconfigured according to equal weight.
Obtaining pattern identifiability F1, wherein the pattern identifiability F1 is the identifiability of the expert identification data, that is, learning the data rule from the data labeled by the expert, obtaining the boundary of the data labeled by the expert by adopting different algorithms, measuring the variance of the accuracy of the XGBOOST algorithm for multiple times, wherein the larger the variance is, the poorer the data quality of the expert label is, the larger the algorithm learning ability changes, the smaller the variance is, calculating the variance of each expert, standardizing the variance to obtain the F1 index of each expert.
The stability of the expert F2 is obtained, the stability of the expert F2 is the consistency of the same data mark of a certain expert at least twice, the higher the stability of the expert is, the deeper the expert knows the standard object, the more stable the mark data is, the higher the reliability of the data quality is.
Acquiring expert service comprehensive capability F3, wherein expert service comprehensive capability F3 refers to expert service comprehensive capability and at least comprises: the level of the expert's business, the title of the expert, the academic calendar of the expert, the working years of the expert, the level of the expert's personality, and the like. The higher the service comprehensive capability of the expert, the higher the quality of the expert data is, and the more reliable the quality of the expert data is.
An expert weight is formed by calculation based on the pattern identifiability F1, the stability of the expert F2 and the expert business integration capability F3.
Each expert reads any client portrait data in the initial client portrait library, wherein the number of client portraits is 34992, and analyzes and judges the current client portrait data to form a judgment result;
each customer portrait data comprises results of M first-type index items and option values of N second-type index items, an expert acquires customer portrait data information from a customer portrait library, reads each option value in each customer portrait data, judges whether a loan request corresponding to the customer portrait data can be passed according to the option values, if the judgment result is 0, the loan request corresponding to the customer portrait data is rejected, and if the judgment result is 1, the loan request corresponding to the customer portrait data is allowed. When the judgment result is 0, the current customer portrait data is used as a critical threshold value, and all customer portrait data lower than the critical threshold value are discarded. And finally, forming a data set with expert decision obtaining determination, wherein in 34992 customer data image combinations, 1000-1500 image combinations need to be judged by an expert. Continuing to form derived data group data by EM-tool algorithm according to the determined data group calculation; where the possible calculated combinations of derived data are approximately 24000. And acquiring deep learning group data by adopting an XGBOOST algorithm based on the determined data group data and the derived data group data, wherein the number of the deep learning group data is about 10000. As shown in the figure, in the above embodiment, it determines the data group data, derives the data group data, and lays out the deep learning group data, wherein the accuracy of the deep learning group data is at least 85%.
In the method, the traditional modeling thought of constructing the expert model by subjectively setting index items and manually distributing weights is abandoned, in order to greatly provide the accuracy of the expert model, a client portrait library is constructed firstly, and then the experts participating in modeling are made to judge only 0 or 1 to form a determined data group. Secondly, 70% of derived data group data is obtained through an EM-tool algorithm based on the determined data group, and finally the remaining 30% of calculation is completed through deep learning of the determined data group data and the derived data group data, so that on one hand, subjectivity of expert judgment is reduced to the lowest, and therefore modeling precision is greatly improved, on the other hand, a large number of experiential creditors are not required to carry out approval, approval efficiency is greatly improved, and meanwhile, approval cost is reduced.
Example two
Based on the first embodiment, this embodiment discloses an application method of a mini-risk control model, and as shown in fig. 6, a flowchart of an application method of a mini-risk control model includes a risk control model formed by any one of the mini-risk control model construction methods provided in the foregoing embodiments, and further includes:
step S201, reading the portrait data to be evaluated of a user, and calculating and obtaining the highest credit limit according to the portrait data to be evaluated and the risk control model;
step S202, approving the loan request of the current user under the condition that the highest loan limit is not less than the target loan limit of the user;
step S203, under the condition that the highest lendable amount is smaller than the target loan amount of the user, forming an advice loan amount according to the highest lendable amount.
Exemplify a specific embodiment
Assume that all the pointers for a certain user are labeled X0, where X0 ═ X1; x 2; x 3; x 4; x 5; x 6; x 7; x 8; x 9; x10], the first category of indicator terms is G0, where G0 ═ x 1; x 2; x 3; x 4; is there a (ii) a x 6; is there a (ii) a Is there a (ii) a x 9; is there a H0, H0 [? (ii) a Is there a (ii) a Is there a (ii) a Is there a (ii) a x 5; is there a (ii) a x 7; x 8; is there a (ii) a x10 ]; and (3) screening all records meeting the G0 combination from the risk control model by taking G0 as a query condition to form a matrix E1 meeting the condition, then deducing the credit amount by 4 boundary thresholds in HO, and then taking the minimum value of the four boundary values as the highest credit amount. And approving the loan requirement of the current user under the condition that the highest lendable amount is not less than the target loan amount of the user, and forming an advice loan amount according to the highest lendable amount under the condition that the highest lendable amount is less than the target loan amount of the user.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (3)
1. A method for constructing a small and micro risk control model is characterized by comprising the following steps: comprises the steps of (a) preparing a mixture of a plurality of raw materials,
screening and forming a sample database in the initial client image library according to a preset method, and forming a determined data group according to the sample database, wherein the method comprises the following steps:
s11, setting M first-class index items and N second-class index items, wherein the first-class index items and the second-class index items are configured with Q option values; forming an initial customer portrait library and portrait combination according to the option values of the first type of index items and the option values of the second type of index items;
step S12, reading any customer portrait data in the initial customer portrait library, and analyzing and judging the current customer portrait data to form a judgment result;
step S13, forming and updating the initial client portrait base according to the judgment result, reading the number of portrait combinations rejected according to the judgment result, and executing step S12 in a state that the number of the rejected portrait combinations is not less than a preset value; otherwise, forming the sample database according to the updated initial client image database;
calculating to form derived data group data according to the determined data group based on EM-tool algorithm;
forming deep learning group data according to the derived data group and the determined data group based on a vector machine algorithm;
the step of reading any client portrait data in the initial client portrait library, and analyzing and judging the current client portrait data to form a judgment result comprises:
step S121, in the state that the current client portrait data is judged to be rejected, a combination pool lower than the current client portrait data is eliminated;
step S122, when the client image data is judged to be approved, accepting a combination pool higher than the current client image data;
the step of forming derived data set data by calculation from the determined data set based on the EM-tool algorithm comprises:
reading a passing rate matched with each option value of a combined line of each portrait data in the determined data set, and acquiring the passing rate of the combined line of the portrait data at present based on the passing rate of each option value;
and sequencing the passing rate of each image data combination line to form the derived data group data.
2. The method for constructing a small risk control model according to claim 1, wherein: forming a deep learning group data according to the derived data group and the determined data group based on a vector machine algorithm comprises the following steps:
reading the determined data group data and the derived data group data, and performing weight processing on the derived data group data to form weight derived data group data;
and performing vector machine processing on the determined data group and the weight derived data group to form the deep learning group data.
3. An application method of a mini risk control model, which is characterized by comprising the risk control model formed based on the mini risk control model construction method of claim 1 or 2, and further comprising:
reading the portrait data to be evaluated of a user, and calculating to obtain the highest credit limit according to the portrait data to be evaluated and the risk control model;
approving the loan request of the current user under the condition that the highest loan limit is not less than the target loan limit of the user;
and forming a suggested loan amount according to the highest lendable amount under the condition that the highest lendable amount is smaller than the target loan amount of the user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810379596.9A CN110400208B (en) | 2018-04-25 | 2018-04-25 | Small and micro risk control model construction method and application method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810379596.9A CN110400208B (en) | 2018-04-25 | 2018-04-25 | Small and micro risk control model construction method and application method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110400208A CN110400208A (en) | 2019-11-01 |
| CN110400208B true CN110400208B (en) | 2021-03-12 |
Family
ID=68319717
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810379596.9A Active CN110400208B (en) | 2018-04-25 | 2018-04-25 | Small and micro risk control model construction method and application method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110400208B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102135979A (en) * | 2010-12-08 | 2011-07-27 | 华为技术有限公司 | Data cleaning method and device |
| CN106447490A (en) * | 2016-09-26 | 2017-02-22 | 广州速鸿信息科技有限公司 | Credit investigation application method based on user figures |
| CN107025596A (en) * | 2016-02-01 | 2017-08-08 | 腾讯科技（深圳）有限公司 | A kind of methods of risk assessment and system |
| CN107798390A (en) * | 2017-11-22 | 2018-03-13 | 阿里巴巴集团控股有限公司 | A kind of training method of machine learning model, device and electronic equipment |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110040631A1 (en) * | 2005-07-09 | 2011-02-17 | Jeffrey Scott Eder | Personalized commerce system |
-
2018
- 2018-04-25 CN CN201810379596.9A patent/CN110400208B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102135979A (en) * | 2010-12-08 | 2011-07-27 | 华为技术有限公司 | Data cleaning method and device |
| CN107025596A (en) * | 2016-02-01 | 2017-08-08 | 腾讯科技（深圳）有限公司 | A kind of methods of risk assessment and system |
| CN106447490A (en) * | 2016-09-26 | 2017-02-22 | 广州速鸿信息科技有限公司 | Credit investigation application method based on user figures |
| CN107798390A (en) * | 2017-11-22 | 2018-03-13 | 阿里巴巴集团控股有限公司 | A kind of training method of machine learning model, device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110400208A (en) | 2019-11-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fisman et al. | Cultural proximity and loan outcomes | |
| Johnes et al. | A comparison of performance of Islamic and conventional banks 2004–2009 | |
| Twati et al. | The impact of organisational culture innovation on the adoption of IS/IT: the case of Libya | |
| Barr et al. | Ethnic fractionalization in an African labour market | |
| US20100268660A1 (en) | Systems and methods for verifying and rating mortgage financial companies | |
| Czaika et al. | High-skilled migration in times of global economic crisis | |
| Wang et al. | Soft information in online peer-to-peer lending: Evidence from a leading platform in China | |
| CN112598500A (en) | Credit processing method and system for non-limit client | |
| CN112541817A (en) | Marketing response processing method and system for potential customers of personal consumption loan | |
| Ereiz | Predicting default loans using machine learning (OptiML) | |
| CN110796539A (en) | Credit investigation evaluation method and device | |
| Chunikhin et al. | Studying the banking industry’s stability through market concentration indices | |
| KR102139938B1 (en) | System for selection of companies subject to credit guarantees based on credit guarantees propensity analysis | |
| Ho | Comparison of Quantitative Shari'ah-Compliant Screening Methods | |
| Muhammad et al. | The determinants of potential failure of islamic peer-to-peer lending: perceptions of stakeholders in Indonesia | |
| Wanke et al. | Revisiting camels rating system and the performance of Asean banks: a comprehensive mcdm/z-numbers approach | |
| Hasan et al. | Financial health and management practices: a multi-year cross country analysis of PLCs | |
| CN112907361A (en) | Method and device for processing loan application | |
| Asikhia | Ethical banking determinants across nations: A meta-analysis | |
| CN110400208B (en) | Small and micro risk control model construction method and application method | |
| CN111191889A (en) | Scoring card development method based on logistic regression and voting type model integration | |
| CN108596755A (en) | The recognition methods of the potential valuable client of bank a kind of and system | |
| CN113837869A (en) | Multi-dimension-based customer analysis and management method | |
| Akram et al. | A club convergence analysis of financial integration: cross-country evidence | |
| CN112734210A (en) | Intelligent case division method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |