CN110263124A - Data detection system - Google Patents
Data detection system Download PDFInfo
- Publication number
- CN110263124A CN110263124A CN201910485351.9A CN201910485351A CN110263124A CN 110263124 A CN110263124 A CN 110263124A CN 201910485351 A CN201910485351 A CN 201910485351A CN 110263124 A CN110263124 A CN 110263124A
- Authority
- CN
- China
- Prior art keywords
- module
- information
- training
- recall rate
- amendment
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
- G06F40/177—Editing, e.g. inserting or deleting of tables; using ruled lines
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of data detection systems, including memory module, integrate module, processing module and sampling module, the processing module includes categorization module, setup module, training module, correction verification module, modified module and test module, and the memory module is stored with historical information；It is described to integrate module integration cleaning carried out to the historical information, generate target histories information；The processing module to the target histories information is classified, be arranged, train, verify and modification obtain amendment training pattern；The sampling module obtains the detection information of object to be detected, and the detection information is input in the amendment training pattern, judging result is obtained.The module of integrating carries out historical information to handle the progress for being conducive to follow-up work, improves efficiency, carries out a series of processing to target histories information in conjunction with the processing module and obtain amendment training pattern, further improve the recall rate of objective attribute target attribute.
Description
Technical field
The present invention relates to technical field of data processing more particularly to a kind of data detection systems.
Background technique
With the quickening of liberalization of international trade process, the trade between various countries exchanges growing, agricultural harmful organism
With Wooden package, worldwide wide-scale distribution and the risk of diffusion also increasingly increase, and need to wooden packing and non-wooden packing
Carry out classification processing.In the prior art, prediction classification is carried out to wooden packing using machine, the data in data model are usually
According to artificial experience by dividing attribute and threshold value being arranged come what is determined, so that wooden packing recall rate only has 1% or so, have very
Big contingency is unfavorable for the classification of wooden packing and non-wood packet.
The Chinese invention patent application of Publication No. CN106447163A discloses a kind of based on fining logical division skill
The quality of data automatic checkout system of art, including the storage of data acquisition module, data categorization module, data evaluation module, data
Module, data correction module, data transmission module, data processing module, database module, abnormal alarm module, data are shown
Module realizes the automatic inspection to the quality of data for successively being classified, assessing, correcting and being handled the initial data of acquisition
It surveys and controls, and stored, shown, exported, effectively improve the quality of data of information.However above-mentioned patent is not to data
Specific classification and appraisal procedure are explained.
Summary of the invention
The purpose of the present invention is to provide a kind of data detection systems, to improve recall rate and accuracy.
To achieve the above object, the data detection system of the invention, including memory module, integrate module, processing mould
Block and sampling module, the processing module include categorization module, setup module, training module, correction verification module, modified module and survey
Die trial block；The memory module storage historical information, archetype, original hyper parameter, recall rate threshold value and amendment recall rate；Institute
It states and integrates module integration cleaning carried out to the historical information, generate target histories information；The categorization module is to the target
Historical information carries out classification processing, to generate training set and test set, and the training set is divided into sub- training set and assessment
Collection；The setup module is configured the original hyper parameter；The training module calls the archetype and setting institute
Original hyper parameter is stated, and processing is trained to the sub- training set with reference to assessment collection, to generate initial training model
And original recall rate；With reference to original recall rate and the recall rate threshold value with to the modified module described in the correction verification module
Modification information is sent, and with reference to the amendment recall rate and the recall rate threshold value to repair to described in test module transmission
Positive training pattern；The modified module modifies processing to the original hyper parameter according to the modification information, to generate
State amendment recall rate and the amendment training pattern；The test module instructs the information input of the test set to the amendment
Practice in model, obtains described with reference to recall rate.The sampling module obtains the detection information of object to be detected, and by the detection
Information input is into the amendment training pattern, to obtain judging result.
The beneficial effects of the present invention are: it is described integrate module to the historical information carry out integration cleaning obtain the mesh
Historical information is marked, the progress that processing is conducive to follow-up work is carried out to the historical information in advance, reduces time loss, improves effect
Rate；The processing module includes the categorization module, the setup module, the training module, the correction verification module, described repairs
Change module and the test module, the processing module is by classifying to the target histories information, being arranged, training, school
It a series of processing such as tests and modifies and obtain the amendment training pattern, improve the accuracy of testing result, improve objective attribute target attribute
Recall rate, while by setting and the hyper parameter is modified, training for promotion efficiency is recalled with the raising amendment training pattern
Rate.
Further, after the correction verification module judges that the original recall rate is less than the recall rate threshold value, Xiang Suoshu is repaired
Change module and sends the modification information.The beneficial effect is that: so that recall rate is met business need.
Further, after the correction verification module judges that the amendment recall rate is more than or equal to the recall rate threshold value, to institute
It states test module and sends the amendment training pattern.The beneficial effect is that: so that recall rate is met business need.
Further, the modified module rolls over cross-validation method with K- and modifies processing to the original hyper parameter, with
Generate the amendment recall rate and amendment training pattern.
Further, the categorization module includes read module, conversion module and segmentation module, and the read module is to institute
It states target histories information to be read out, the conversion module converts the target histories information of reading, forms target
History data set, the segmentation module are split to form training set and test set to the target histories data set, then will
The training set is divided into sub- training set and assessment collection.
Further, the read module is read out the target histories information with Pandas software, the conversion
The target histories information is converted to the data set of DateFrame structure by module, and the segmentation module is promoted with gradient
Decision tree open source realizes that the training set is divided into the sub- training set by software and the assessment collects.
Further, the recall rate threshold value is greater than 0.8, and the recall rate threshold value is to measure the index of model performance, industry
Recall rate is required to be greater than 0.8 in business.
Further, the memory module is also stored with weight, and the training module is by calling the archetype, institute
Original hyper parameter and the weight are stated, and processing is trained to the sub- training set with reference to assessment collection, generates initial instruction
Practice model and original recall rate.The beneficial effect is that: recall rate can be improved in the weight.
Further, the module of integrating includes relating module and cleaning module, and the relating module believes the history
Breath is associated, and the cleaning module cleans the historical information, obtains the target histories information.Its advantages
It is: obtains the characteristic data items of training pattern needs.
Further, the historical information includes total waybill database table, House Air Waybill database table and attribute record table.
Further, the relating module is used to carry out the House Air Waybill database table and total waybill database table
External key association, and the external key association is carried out to the House Air Waybill database table and the attribute record table.
Further, the attribute record table is record sheet containing wooden packing.
Further, the cleaning module cleans total waybill database table and the House Air Waybill database table
Obtain all total waybill records and the partite transport unirecord comprising wooden packing.
Further, the detection information be total waybill goods weight, total waybill cargo number of packages, House Air Waybill goods weight,
One or more of House Air Waybill cargo number of packages, House Air Waybill value of goods, bill of lading number and currency type.The beneficial effect is that:
The accuracy of judging result is improved by different detection informations.
Detailed description of the invention
Fig. 1 is the structural block diagram of data detection system of the invention；
Fig. 2 is the structural block diagram for integrating module of the invention；
Fig. 3 is the structural block diagram of categorization module of the invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing of the invention, to this hair
Technical solution in bright embodiment is clearly and completely described, it is clear that described embodiment is that a part of the invention is implemented
Example, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creativeness
Every other embodiment obtained, shall fall within the protection scope of the present invention under the premise of labour.Unless otherwise defined, it is used herein as
Technical term or scientific term should be the usual meaning that persons with general skills in the field are understood
Justice.The similar word such as " comprising " used herein, which means to occur element or object before the word, to be covered and appears in the word
The element of presented hereinafter perhaps object and its equivalent and be not excluded for other elements or object.
In view of the problems of the existing technology, some embodiments of the present invention provide a kind of data detection system, have
Memory module integrates module, processing module and sampling module.
In some embodiments of the present invention, the data detection system setting is on computers.
Fig. 1 is a kind of structural block diagram of data detection system of the invention.Referring to Fig.1, data detection system 1 has storage
Module 11 integrates module 12, processing module 13 and sampling module 14, and the processing module 13 has categorization module 131, setting mould
Block 132, training module 133, correction verification module 134, modified module 135 and test module 136.The memory module 11, which is stored with, to be gone through
History information, archetype, original hyper parameter, recall rate threshold value, with reference to recall rate, weight and table on exchange rates.It is described to integrate module 12
Integration cleaning treatment is carried out to the historical information, generates target histories information；The processing module 13 is to the target histories
Information is modified processing, obtains amendment training pattern；The sampling module 14 is used to obtain the detection information of object to be detected, with
And the detection information is input in the amendment training pattern, to obtain judging result.
In some specific embodiments of the present invention, the historical information is total waybill database table, House Air Waybill database table
With attribute record table, the attribute in the attribute record table is divided into objective attribute target attribute and characteristic attribute, and the objective attribute target attribute is wood packet
Dress, the characteristic attribute be total waybill goods weight, total waybill cargo number of packages, House Air Waybill goods weight, House Air Waybill cargo number of packages,
House Air Waybill value of goods, bill of lading number and the exchange rate.
In particular embodiments of the invention, the detection information is House Air Waybill goods weight, House Air Waybill value, the bill of lading
One or more of number, House Air Waybill number of packages, total waybill cargo number of packages, total waybill goods weight and exchange rate believe the detection
Breath is input in the amendment training pattern, obtains judging result, if is wooden packing.
With reference to Fig. 1, the categorization module 131 is to the target histories information classification processing.The categorization module 131 is on time
Between sequence carry out the target histories data information to be divided into training set and test set, further, the categorization module 131 will
The training set is divided into sub- training set and assessment collection.The training set accounts for the target histories information in certain embodiments
70%, the test set accounts for the 30% of the target histories information, and the training set is divided into ten by the categorization module 131
Part, wherein nine parts are sub- training set, remaining a to assess collection, i.e., the described sub- training set is the target histories data
63%, the assessment collection is the 7% of the target histories data.It is more specifically to promote decision tree using a kind of gradient and increase income
Realize that the training set is divided into ten parts by software, the gradient promotes decision tree open source and realizes that software is LightGBM.
The original hyper parameter and the weight is arranged in the setup module 132.
The training module 133 refers to institute by calling the archetype, the original hyper parameter and the weight
Commentary estimates collection and is trained processing to the sub- training set, generates initial training model and original recall rate, evaluation index are
Recall rate.
The original hyper parameter all uses the default value of the archetype itself.
Since wooden packing proportion is about 5% or so in practical business, belong to class imbalance problem, is needed when training
Weight is arranged to record, to improve recall rate and accuracy rate.
The weight is divided into wooden packing weight and non-wooden packing weight, the calculation formula of the wooden packing weight are as follows: wood packet
Fill weight=training set record number/(wooden packing records number in 2 × training set)；The calculation formula of the non-wooden packing weight are as follows:
Non- wooden packing weight=training set record number/(non-wooden packing records number in 2 × training set).
The original hyper parameter includes the number of iterations and learning rate, with gradually increasing for the number of iterations, the original
The selection range of beginning hyper parameter is gradually reduced, and the training pattern is more and more accurate, and the preferred the number of iterations is 20 times, institute
Stating learning rate is a coefficient, is equivalent to the minimum of slope, and the best learning rate is 0.5, modifies the number of iterations energy
Enough training for promotion efficiency, modifying the learning rate can be improved the recall rate of model.
In the training process, sub- training set described in the training module has wooden packing data and non-wood packaged data two
Kind situation, if the truth of sub- training set is with N number of wooden packing data and M non-wood packaged datas, wherein N number of wood
The prediction case of packaged data is that the quantity of wooden packing is TP and the quantity of non-wooden packing is FN, the calculation formula of recall rate are as follows:
Recall rate=TP/ (TP+FN)；The prediction case of the M non-wood packaged datas is that the quantity of wooden packing is FP and non-wooden packing
Quantity be TN, the calculation formula of accuracy rate are as follows: accuracy rate=TP/ (TP+FP).
Some specific trainings, the truth of sub- training set are 46641 wooden packing data, prediction case: TP
=38689, FN=7952；The truth of sub- training set be 770377 non-wood packaged datas, prediction case: FP=49798,
TN=720579.Accordingly wherein recall rate are as follows: recall rate=38689/ (38689+7952)=0.83；Accuracy rate
Are as follows: accuracy rate=38689/ (38689+49798)=0.4.
Processing is compared with recall rate threshold value in the original recall rate by the correction verification module 134, when judging the original
After beginning recall rate is less than the recall rate threshold value, Xiang Suoshu modified module 135 sends modification information.The recall rate threshold value=
0.8。
The modified module 135 modifies processing to the original hyper parameter according to the modification information, generates amendment
Recall rate and amendment training pattern.The K- of modified module 135 described in some specific embodiments rolls over cross-validation method to described
Original hyper parameter is modified processing, to generate the amendment recall rate and amendment training pattern.
The training module 133 calls the amendment recall rate and the amendment training pattern, repeat before the step of it is straight
Judge that the amendment recall rate is more than or equal to the recall rate threshold value to the correction verification module 134, that is, it is raw after being more than or equal to 0.8
At amendment training pattern, Xiang Suoshu test module 136 sends the amendment training pattern.
The information input of the test set into the amendment training pattern, is obtained the ginseng by the test module 136
Examine recall rate and final accuracy rate.
In some specific embodiments, the archetype is a kind of integrated model based on decision tree, the training
The record containing wooden packing and without containing wooden packing is split up into target, each calculates step and selects an input special module
Xi is levied, the correction verification module establishes a rule in the corresponding cut-point si of the feature Xi, and the rule includes to original super
Whether the setting of parameter, weight and recall rate threshold value, the regular data characteristics attribute for judging input contain wooden packing,
Need to carry out next calculating step for the rule of classification error, i.e., the described modified module is special by reselecting new input
Cut-point of seeking peace further is divided, until reaching preset calculating step or recording completely separable by all, establishes one
Decision tree, i.e. amendment training pattern.
Fig. 2 is the structural block diagram for integrating module of the invention, and with reference to Fig. 2, the module 12 of integrating is with relating module 121
With cleaning module 122, the relating module 121 is to total waybill database table, the House Air Waybill database table and described
Attribute record table carries out external key association, and the cleaning module 122 is to total waybill database table, the House Air Waybill database table
And the attribute record table carries out integration cleaning treatment, obtains target histories information.It is described in some specific embodiments
Attribute record table is Excel format.The cleaning module 122 described in some specific embodiments to external key be associated with after described in
Total waybill database table, the House Air Waybill database table and the attribute record table start the cleaning processing, and obtain target histories
Information.The target histories information is stored in the file of parquet format by memory module described in some specific embodiments.
Fig. 3 is the structural block diagram of categorization module of the invention, and with reference to Fig. 3, the categorization module 131 has read module
1311, conversion module 1312 and segmentation module 1313, the read module 1311 read the target histories data, while described
Conversion module 1312 converts the target histories information to form target histories data set, and the segmentation module 1313 is by the mesh
Mark history data set is divided into the training set and the test set, and the training set is divided into the sub- training set and institute's commentary
Estimate collection.
In some specific embodiments, the read module 1311 reads target histories information with data processing software, more
Specially target histories information is read with Pandas；Conversion module 1312 described in some specific embodiments is by the target histories
Information is converted to the data set of DateFrame structure.
Although embodiments of the present invention are hereinbefore described in detail, show for those skilled in the art
And be clear to, these embodiments can be carry out various modifications and be changed.However, it is understood that this modifications and variations are all
Belong within scope and spirit of the present invention described in the claims.Moreover, the present invention described herein can have others
Embodiment, and can be practiced or carried out in several ways.
Claims (14)
1. a kind of data detection system, which is characterized in that including memory module, integrate module, processing module and sampling module, institute
Stating processing module includes categorization module, setup module, training module, correction verification module, modified module and test module；
The memory module storage historical information, archetype, original hyper parameter, recall rate threshold value and amendment recall rate；
It is described to integrate module integration cleaning treatment carried out to the historical information, generate target histories information；
The categorization module carries out classification processing to the target histories information, to generate training set and test set, and by institute
It states training set and is divided into sub- training set and assessment collection；
The setup module is configured the original hyper parameter；
The training module calls the archetype and the original hyper parameter is arranged, and collects with reference to the assessment to described
Sub- training set is trained processing, to generate initial training model and original recall rate；
The correction verification module is with reference to the original recall rate and the recall rate threshold value to send modification letter to the modified module
Breath, and with reference to the amendment recall rate and the recall rate threshold value to send amendment training pattern to the test module；
The modified module modifies processing to the original hyper parameter according to the modification information, is called together with generating the amendment
Return rate and the amendment training pattern；
The information input of the test set into the amendment training pattern, is obtained the reference and recalled by the test module
Rate；
The sampling module obtains the detection information of object to be detected, and the detection information is input to the amendment training mould
In type, to obtain judging result.
2. data detection system according to claim 1, which is characterized in that the correction verification module judges described original recall
After rate is less than the recall rate threshold value, Xiang Suoshu modified module sends the modification information.
3. data detection system according to claim 1, which is characterized in that the correction verification module judges that the amendment is recalled
After rate is more than or equal to the recall rate threshold value, Xiang Suoshu test module sends the amendment training pattern.
4. data detection system according to claim 1, which is characterized in that the modified module rolls over cross-validation method with K-
It modifies processing to the original hyper parameter, to generate the amendment recall rate and amendment training pattern.
5. the data detection system according to claim 1, which is characterized in that the categorization module includes reading mould
Block, conversion module and segmentation module, the read module are read out the target histories information, and the conversion module is to reading
The target histories information taken is converted, and forms target histories data set, the segmentation module is to the target histories number
It is split to form training set and test set according to collection, and the training set is divided into sub- training set and assessment collection.
6. the data detection system according to claim 5, which is characterized in that read module Pandas software
The target histories information is read out, the target histories information is converted to DateFrame structure by the conversion module
The data set, the segmentation module promotes decision tree open source with gradient and realizes that the training set is divided into the sub- instruction by software
Practice collection and assessment collection.
7. data detection system according to claim 1, which is characterized in that the recall rate threshold value is more than or equal to 0.8.
8. data detection system according to claim 1, which is characterized in that the memory module is also stored with weight, institute
Training module is stated to collect by calling the archetype, the original hyper parameter and the weight, and with reference to the assessment to institute
It states sub- training set and is trained processing, generate initial training model and original recall rate.
9. data detection system according to claim 1, which is characterized in that the module of integrating includes relating module and clear
Mold cleaning block, the relating module are associated the historical information, and the cleaning module cleans the historical information,
Obtain the target histories information.
10. data detection system according to claim 9, which is characterized in that the historical information includes total waybill data
Library table, House Air Waybill database table and attribute record table.
11. data detection system according to claim 10, which is characterized in that the relating module is to the partite transport odd number
External key association is carried out according to Ku Biao and total waybill database table, and to the House Air Waybill database table and the attribute record
Table carries out the external key association.
12. data detection system according to claim 10, which is characterized in that the attribute record table is to remember containing wooden packing
Record table.
13. data detection system according to claim 10, which is characterized in that the cleaning module is to total waybill number
Cleaned to obtain all total waybill records and partite transport unirecord comprising wooden packing according to Ku Biao and the House Air Waybill database table.
14. data detection system according to claim 1, which is characterized in that the detection information is total waybill cargo weight
Amount, total waybill cargo number of packages, House Air Waybill goods weight, House Air Waybill cargo number of packages, House Air Waybill value of goods, bill of lading number and currency class
One or more of type.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201821962014 | 2018-11-27 | ||
| CN2018219620141 | 2018-11-27 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110263124A true CN110263124A (en) | 2019-09-20 |
Family
ID=67916953
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910485351.9A Pending CN110263124A (en) | 2018-11-27 | 2019-06-05 | Data detection system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110263124A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529565A (en) * | 2016-09-23 | 2017-03-22 | 北京市商汤科技开发有限公司 | Target identification model training and target identification method and device, and computing equipment |
| US20180240041A1 (en) * | 2017-02-22 | 2018-08-23 | Sas Institute Inc. | Distributed hyperparameter tuning system for machine learning |
| CN108537289A (en) * | 2018-04-24 | 2018-09-14 | 百度在线网络技术（北京）有限公司 | Training method, device and the storage medium of data identification model |
-
2019
- 2019-06-05 CN CN201910485351.9A patent/CN110263124A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529565A (en) * | 2016-09-23 | 2017-03-22 | 北京市商汤科技开发有限公司 | Target identification model training and target identification method and device, and computing equipment |
| US20180240041A1 (en) * | 2017-02-22 | 2018-08-23 | Sas Institute Inc. | Distributed hyperparameter tuning system for machine learning |
| CN108537289A (en) * | 2018-04-24 | 2018-09-14 | 百度在线网络技术（北京）有限公司 | Training method, device and the storage medium of data identification model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109360084A (en) | Appraisal procedure and device, storage medium, the computer equipment of reference default risk | |
| CN106156809A (en) | For updating the method and device of disaggregated model | |
| CN106874658A (en) | A kind of medical insurance fraud recognition methods based on Principal Component Analysis Algorithm | |
| KR100811307B1 (en) | Prediction method for earning rate in construction | |
| CN107545038A (en) | A kind of file classification method and equipment | |
| CN111932044A (en) | Steel product price prediction system and method based on machine learning | |
| CN109741175A (en) | Based on artificial intelligence to the appraisal procedure of credit again and equipment for purchasing automobile-used family by stages | |
| CN113159461A (en) | Small and medium-sized micro-enterprise credit evaluation method based on sample transfer learning | |
| CN114419397A (en) | Data set construction method and device based on data cleaning and data generation | |
| CN113240272A (en) | Enterprise ESG index determination method and related product | |
| CN107862599B (en) | Bank risk data processing method and device, computer equipment and storage medium | |
| CN111738870A (en) | Method and platform for identifying insurance risk of engineering performance guarantee based on characteristic engineering | |
| CN110263124A (en) | Data detection system | |
| CN112037006A (en) | Credit risk identification method and device for small and micro enterprises | |
| KR101536536B1 (en) | Method for evaluating technology by adjusted reproduction cost approaching | |
| CN114781948B (en) | Data processing method and related device for determining working quality of staff | |
| CN115829722A (en) | Training method of credit risk scoring model and credit risk scoring method | |
| CN105488521B (en) | A kind of dilatation screening sample method based on kernel function | |
| CN115239502A (en) | Analyst simulation method, analyst simulation system, electronic device and storage medium | |
| CN112884301A (en) | Method, equipment and computer storage medium for enterprise risk analysis | |
| WANG et al. | Research on Credit Decision Issues of the Small and Medium-Sized Enterprises Based on TOPSIS and Hierarchical Cluster Analysis [C] | |
| CN116933187A (en) | Digital processing method and system for financial reimbursement bill | |
| CN117853225A (en) | Credit evaluation method for debt subject | |
| CN116629652A (en) | Code quality and personal contribution evaluation method for multi-person collaboration project | |
| CN109472492A (en) | A kind of performance quantization assessment system and method |
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 |