CN115829205A - Method and device for monitoring due-employment level of manual credit examiner for assisting credit of small and micro enterprises - Google Patents
Method and device for monitoring due-employment level of manual credit examiner for assisting credit of small and micro enterprises Download PDFInfo
- Publication number
- CN115829205A CN115829205A CN202211493249.1A CN202211493249A CN115829205A CN 115829205 A CN115829205 A CN 115829205A CN 202211493249 A CN202211493249 A CN 202211493249A CN 115829205 A CN115829205 A CN 115829205A
- Authority
- CN
- China
- Prior art keywords
- credit
- information
- model
- training
- monitoring
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000012544 monitoring process Methods 0.000 title claims abstract description 23
- 239000000463 material Substances 0.000 claims abstract description 51
- 230000006399 behavior Effects 0.000 claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 34
- 239000013598 vector Substances 0.000 claims abstract description 15
- 238000012545 processing Methods 0.000 claims abstract description 9
- 238000002372 labelling Methods 0.000 claims abstract description 5
- 238000001514 detection method Methods 0.000 claims description 11
- 206010000117 Abnormal behaviour Diseases 0.000 claims description 6
- 238000012216 screening Methods 0.000 claims description 6
- 230000005856 abnormality Effects 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000012706 support-vector machine Methods 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 abstract 1
- 238000004519 manufacturing process Methods 0.000 abstract 1
- 238000012550 audit Methods 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of machine learning, and discloses a method and a device for monitoring the full-time level of a manual credit reviewer for assisting credit of a small enterprise, wherein S1, material information to be checked by the credit reviewer is subjected to standardization processing to manufacture content links of tables or graphs, the content links are stored and displayed in a form of title and cover, and finally a content material library is formed; and S2, collecting related characteristic information, processing the characteristic information into a standardized characteristic vector X, labeling paths of materials actually used by an approver to form a prediction target Y, and using the characteristic X and the target Y as input of recommendation model training. The method and the device for monitoring the full-time level of the manual credit examiner assisting the credit of the small and micro enterprise can extract the information of the object to be examined, make readable materials, intelligently recommend the materials to the user, record the behaviors of the user, realize transverse and longitudinal comparison of the behaviors of all the users in the system and achieve the purpose of monitoring the behavior specification and abnormity.
Description
Technical Field
The invention relates to the technical field of communication, in particular to a method and a device for monitoring the due-employment level of an artificial creditor assisting the credit of a small and micro enterprise.
Background
At present, the financial technology realizes the online process of applying for loan by a client, and some standardized links such as admission condition screening have realized automatic decision-making based on client score distribution, but the part with high subjective judgment requirement needs the participation of manual examination and approval personnel, and finally makes a comprehensive decision of whether to give credit, the amount of the given credit and the rate of the given credit based on industry experience, market preference, the reasonability of fund application and the reliability of repayment sources and the like. Because the industry span of small and micro enterprises is large, the data standardization degree is low, the decision-making right of the auditors has larger degree of freedom, and whether the auditors work or not is responsible for the credit assets, which is particularly important for the safety of the credit assets, a set of method and a system for monitoring the full-time degree of manual credit audit are needed.
When evaluating an enterprise, a credit examiner needs to refer to multidimensional information, such as applicant credit history, enterprise asset liability, enterprise associated network information, social public opinion information, market popularity of the enterprise belonging to the industry, market period of the market and the like, a large amount of information needs to be searched and consulted, sometimes even repeatedly compared, and the question of one information point triggers the search of the next information point. The method in the market at present is to record the collected information into a database through an approval system, provide a query interface, support the development of a customized report system, display the information required by letter and audit personnel, but lack the function of intelligently analyzing and monitoring the full-time level of the letter and audit personnel.
The defects in the prior art are as follows: the monitoring of the overdue rate of the loan can be shown after the actual occurrence of the risk, so that the problem is found late, the damage is not stopped as early as possible, the risk strategy is adjusted, and the monitoring means lacks the self-learning updating capability.
Disclosure of Invention
Aiming at the problems or the defects, the invention provides the method and the device for monitoring the full-time level of the manual credit examiner for assisting the credit of the small and micro enterprises, which can extract the information of the object to be examined, make readable materials, intelligently recommend the materials to the user, record the behaviors of the user, realize the transverse and longitudinal comparison of all the behaviors of the user in the system and achieve the purpose of monitoring the behavior specification and abnormity.
In order to achieve the purpose, the invention adopts the following technical scheme:
the method and the device for monitoring the due-employment level of the manual credit examiner for assisting the credit of the small and micro enterprises comprise the following steps:
s1, standardizing material information to be checked by letter examiners, making content links of tables or graphs, storing and displaying the content links in a title and cover page mode, and finally forming a content material library;
s2, collecting related characteristic information, processing the characteristic information into a standardized characteristic vector X, labeling paths of materials actually used by an approver to form a prediction target Y, and using the characteristic X and the target Y as input of recommendation model training;
s3, a GBDT + LR combined model is adopted by a training model, GBDT is a method for automatically screening features and combined features, discrete feature vectors can be generated, then the discrete feature vectors are used as input, the click rate of a user on recommended contents is predicted by an LR model, and then the contents ranked in the front are screened to output recommended materials;
and S4, training the network relation of all users by adopting a training model to obtain an abnormality detection model, and predicting the due-employment level by judging the deviation degree of the examination and approval personnel and the generalized behavior.
Preferably, the material information in S1 includes: the system comprises applicant-related relationship information, applicant main enterprise related information, enterprise affiliated industry information and applicant or enterprise main network public opinion information.
Preferably, the related feature information in S2 includes: behavior data of the approver, environmental information and characteristics of the checked object.
Preferably, the network relationship of S4 is defined as follows: when using the recommended materials, the approver generates a large amount of behavior records, called behavior paths, which form a network relationship between the user and the selected materials.
Preferably, the training model in S4 is a deep hybrid model, the deep hybrid model performs feature extraction using a self-encoder, and the extracted features are input to a one-class support vector machine for training.
Preferably, after model training, the behavior track of the examiner can calculate a statistic of quantitative difference on data characteristics, the statistic is called distance, normal behaviors are gathered together, and the farther the behavior track is from the center point of the normal behaviors, the abnormal behaviors are judged.
In order to achieve the purpose, the invention also adopts the following technical scheme:
the method and the device for monitoring the due-employment level of the manual credit examiner for assisting the credit of the small and micro enterprises comprise a content material making module, a training recommendation system module and a system abnormity detection model module; wherein,
the content material making module is used for carrying out standardized processing on material information to be checked by a creditor, making content links of tables or graphs, storing and displaying the content links in a title and cover page mode and finally forming a content material library;
the training recommendation system module is used for collecting relevant feature information, processing the feature information into a standardized feature vector X, labeling paths of materials actually used by the examination and approval personnel to form a prediction target Y, and using the feature X and the target Y as input of the training of a recommendation model; the training model adopts a GBDT + LR combined model, GBDT is a method for automatically screening features and combining features, discrete feature vectors can be generated, then the discrete feature vectors are used as input, the click rate of a user on recommended contents is predicted by the LR model, and then the contents ranked at the front are screened out to output recommended materials;
and the system anomaly detection model module is used for obtaining an anomaly detection model and predicting the due-job level by judging the deviation degree of the examining and approving personnel from the generalized behaviors.
The invention has the beneficial effects that:
the invention provides a method and a device for monitoring the full-time level of a manual credit examiner for assisting the credit of a small and micro enterprise, which can extract the information of an object to be examined, make a readable material, intelligently recommend the material to a user, record the behavior of the user, realize transverse and longitudinal comparison of all the user behaviors in a system and achieve the purposes of monitoring behavior specification and abnormity; the basic implementation scheme is that the client information and the associated enterprise information are processed in a standardized manner to be made into table or graphic materials, the system can recommend the materials to an approver, and intelligent recommendation is formed by learning the behaviors of the approver in using the materials, such as retention time, click rate, sequence for viewing the materials, subscription information and the like, so that the approval efficiency is greatly improved, and meanwhile, the behaviors of a user are monitored, and the full-time level of the user is evaluated in real time.
Drawings
FIG. 1 is a flow chart of a recommendation system training of the present invention;
FIG. 2 is a flow chart of the abnormal behavior detection training of the present invention;
FIG. 3 is a flow chart of an embodiment of the present invention.
Detailed Description
The technical method of the invention is described in detail in the following with reference to the accompanying drawings of one embodiment of the invention. It is to be understood that the embodiments described are only a few embodiments of the present application and not all embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application.
In the description of the embodiments of the present application, it should be noted that the terms "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings or orientations or positional relationships that the products of the present invention are usually placed in when used, and are only used for convenience of description and simplicity of description, but do not indicate or imply that the devices or elements that are referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present application. Furthermore, the terms "first," "second," "third," and the like are used solely to distinguish one from another and are not to be construed as indicating or implying relative importance.
The present invention will be described in detail with reference to the drawings
The implementation steps are as follows, as shown in fig. 3:
step 1: producing content material
Materials needing to be checked by the creditor comprise 1) applicant-related relationship information; 2) Applicant subject business related information; 3) Associated enterprise related information of the subject enterprise; 4) Information of industries to which enterprises belong; 5) The applicant or the enterprise main body network public opinion information (websites, public numbers, search engines and the like) is standardized, made into content links of tables or graphs and stored and displayed in the form of titles and covers. Such as revenue, profit, gross profit margin, ROE, and ROA, titled enterprise's last year quarter 1. The approver can click the cover to view the corresponding index content, and browse the next material in a screen refreshing or next step clicking mode.
Step 2: processing features
The features herein refer to:
1) Behavior data of the approver (for example, the click rate, the stay time, the click sequence and other browsing records of the approver on the content material once);
2) Environmental information (such as login time, current approval node position, behavior data of other approval personnel and the like);
3) Characteristics of the object under review (e.g.: industry of the business, size of the business, location of the business, type of business, etc.).
The approval system can collect the above characteristics, process the information into a standardized characteristic vector X, and label the path of the material actually used by the approver to form a prediction target Y (for example, under the condition that the characteristic is X, the approver actually makes the path for looking through the content material be A- > B- > C- > D- > E, namely, firstly looks over the A material, then looks over the B material …, and finally makes an approval resolution after finishing looking over the E material), and the characteristic X and the target Y are used as the input of the recommendation model training.
And step 3: training recommendation system
The core task of the recommendation algorithm is to screen out the most favorite options from massive material contents under the characteristics of a given user and continuously learn to optimize the recommendation accuracy. The training model of the invention adopts a combined model of GBDT + LR (gradient lifting decision tree + logistic regression), GBDT is a method for automatically screening features and combined features, discrete feature vectors can be generated, then the click rate CTR of a user on recommended contents is predicted by the logistic regression model by taking the discrete features as input, and then contents ranked at the top (generally top 10) are screened and output. The combined model has the advantages of obtaining the advantages of feature combination, controlling the complexity of the model and reducing the requirement on the operation performance. The training is divided into an off-line model training part and an on-line prediction output part, as shown in fig. 1.
Step 4, training and approving abnormal model
When the approving personnel uses the intelligently recommended materials, a large number of behavior records are generated, called behavior paths, and the paths form the network relationship between the User and the selected materials (for example, the approving personnel User0 checks the income of tax declaration of an enterprise, the history of credit investigation and liability of the enterprise, the total amount of personal credit investigation and liability, the inquiry record of recently applied credit, the court litigation cases related to the enterprise, then jumps to the core index of the asset liability table of tax declaration, finally makes an approval decision, and the stay time of each node is also recorded). Training the network of all users (including a small number of abnormal behavior labels), obtaining an abnormal detection model, and predicting due-time level by judging the deviation degree of the examining and approving personnel from the generalized behavior. The behavior track of the examiner after model training can calculate the statistic of the quantitative difference on the data characteristics, so that the statistic can be called as the distance to be conveniently understood, normal behaviors are gathered together, and the farther the behavior track is from the central point of the normal behavior, the more easily the behavior track is judged to be abnormal behavior, for example: the normal behaviors mostly fall near the origin (0,0,0) of the three-dimensional coordinate system, and a specific behavior (100,100,100) is seriously deviated from the origin to be judged as abnormal behavior.
The invention is trained using a deep hybrid model. The deep hybrid model uses an Auto-Encoder (Auto-Encoder) for feature extraction, and the extracted features are sent to a single-class support vector machine (OC-SVM) of the traditional anomaly detection algorithm for training as shown in FIG. 2.
The above-mentioned embodiments only express the specific embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for those skilled in the art, without departing from the technical idea of the present application, several changes and modifications can be made, which are all within the protection scope of the present application.
Claims (7)
1. The method and the device for monitoring the due-employment level of the manual credit examiner for assisting the credit of the small and micro enterprises are characterized by comprising the following steps of:
s1, standardizing material information to be checked by letter examiners, making content links of tables or graphs, storing and displaying the content links in a title and cover page mode, and finally forming a content material library;
s2, collecting related characteristic information, processing the characteristic information into a standardized characteristic vector X, labeling paths of materials actually used by an approver to form a prediction target Y, and using the characteristic X and the target Y as input of recommendation model training;
s3, a GBDT + LR combined model is adopted by a training model, GBDT is a method for automatically screening features and combined features, discrete feature vectors can be generated, then the discrete feature vectors are used as input, the click rate of a user on recommended contents is predicted by an LR model, and then the contents ranked in the front are screened to output recommended materials;
and S4, training the network relation of all users by adopting a training model to obtain an abnormality detection model, and predicting the due-employment level by judging the deviation degree of the examination and approval personnel and the generalized behavior.
2. The method and apparatus for monitoring the due diligence level of human credit assistance on small business as claimed in claim 1, wherein the material information in S1 includes: the system comprises applicant related relation information, applicant main body enterprise related information, related enterprise related information of main body enterprises, industry information of enterprises, and applicant or enterprise main body network public opinion information.
3. The method and apparatus for assisting in the monitoring of the due diligence level of credit for small business as claimed in claim 1, wherein the related characteristic information of S2 includes: the behavior data of the examiners, the environmental information and the characteristics of the examined objects.
4. The method and apparatus for monitoring the due diligence level of human credit assistance for small business enterprises as claimed in claim 1, wherein the network relationship of S4 is defined as follows: when using the recommended materials, the approver generates a large amount of behavior records, called behavior paths, which form a network relationship between the user and the selected materials.
5. The method and apparatus for manual credit reviewer full-time level monitoring with assistance to small business credit as claimed in claim 4, wherein the training model in S4 is a deep hybrid model, the deep hybrid model uses a self-encoder for feature extraction, and the extracted features are used as input to a one-class support vector machine for training.
6. The method and apparatus for monitoring the due diligence level of credit assistance for small business enterprises as claimed in claim 5, wherein the behavior trajectory of the approver after model training can calculate the statistic of quantitative difference on the data characteristics, and the statistic is called distance, the normal behaviors are gathered together, and the farther from the center point of the normal behaviors, the abnormal behaviors are determined.
7. The method and apparatus for monitoring the due diligence level of human credit assistance for small business enterprises as claimed in claim 1, comprising a content material making module, a training recommendation system module and a system abnormality detection model module; wherein,
the content material making module is used for carrying out standardized processing on material information to be checked by a creditor, making content links of tables or graphs, storing and displaying the content links in a title and cover page mode and finally forming a content material library;
the training recommendation system module is used for collecting relevant feature information, processing the feature information into a standardized feature vector X, labeling paths of materials actually used by the examination and approval personnel to form a prediction target Y, and using the feature X and the target Y as input of the training of a recommendation model; the training model adopts a GBDT + LR combined model, GBDT is a method for automatically screening features and combined features, discrete feature vectors can be generated, then the discrete feature vectors are used as input, the click rate of a user on recommended contents is predicted by an LR model, and then the contents ranked in the front are screened to output recommended materials;
and the system anomaly detection model module is used for obtaining an anomaly detection model and predicting the due-job level by judging the deviation degree of the examining and approving personnel from the generalized behaviors.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211493249.1A CN115829205A (en) | 2022-11-25 | 2022-11-25 | Method and device for monitoring due-employment level of manual credit examiner for assisting credit of small and micro enterprises |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211493249.1A CN115829205A (en) | 2022-11-25 | 2022-11-25 | Method and device for monitoring due-employment level of manual credit examiner for assisting credit of small and micro enterprises |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115829205A true CN115829205A (en) | 2023-03-21 |
Family
ID=85531666
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211493249.1A Pending CN115829205A (en) | 2022-11-25 | 2022-11-25 | Method and device for monitoring due-employment level of manual credit examiner for assisting credit of small and micro enterprises |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115829205A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN118279048A (en) * | 2024-06-03 | 2024-07-02 | 杭银消费金融股份有限公司 | Automatic generation method and system for wind control strategy |
-
2022
- 2022-11-25 CN CN202211493249.1A patent/CN115829205A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN118279048A (en) * | 2024-06-03 | 2024-07-02 | 杭银消费金融股份有限公司 | Automatic generation method and system for wind control strategy |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kara et al. | A data mining-based framework for supply chain risk management | |
Bachmann et al. | The impact of international outsourcing on labour market dynamics in Germany | |
Khalifa et al. | Exploratory research on digitalization transformation practices within supply chain management context in developing countries specifically Egypt in the MENA region | |
Lee et al. | Fundamentals of big data network analysis for research and industry | |
CN115619571A (en) | Financing planning method, system and device | |
CN114528416B (en) | Enterprise public opinion environment monitoring method and system based on big data | |
CN116823498A (en) | Personalized insurance product recommendation engine system based on big data analysis and machine learning | |
CN111951093A (en) | Personal credit score scoring method | |
CN115829205A (en) | Method and device for monitoring due-employment level of manual credit examiner for assisting credit of small and micro enterprises | |
CN114299350B (en) | Manual credit auditing information recommendation method and system based on machine learning | |
Nikolić et al. | Application of FAHP–PROMETHEE hybrid model for prioritizing SMEs failure factors | |
KR102499182B1 (en) | Loan regular auditing system using artificia intellicence | |
CN113705072A (en) | Data processing method, data processing device, computer equipment and storage medium | |
US20200342302A1 (en) | Cognitive forecasting | |
Daniel et al. | Developing a new multidimensional model for selecting strategic plans in balanced scorecard | |
Ayyasamy et al. | Design and implementation of residential rental rates forecast model using data mining algorithms | |
Murugan et al. | Sales Forecasting using SARIMAX for B2C | |
Haverila et al. | Mapping the marketing analytics landscape: unveiling the diverse proficiencies in marketing and technical know-how among practitioners | |
Bettahar et al. | The role of big data analysis and strategic vigilance in decision-making | |
LU505415B1 (en) | System and method for pushing digital financial businesses based on artifical intelligence | |
Bai | The application of customer relationship management and data mining in Chinese insurance companies | |
Kallio | A multi-case study of the analytical capabilities of Finnish e-commerce businesses: a resource-based view | |
OgrajenSek | Use of customer data analysis in continuous quality improvement of service processes | |
Punithadevi et al. | Analyzing students academic performance using internal factors—A survey | |
Abdon et al. | Big Data for Small, Midsize, and Large Operations |
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 |