CN110414716A - A kind of enterprise based on LightGBM breaks one's promise probability forecasting method and system - Google Patents
A kind of enterprise based on LightGBM breaks one's promise probability forecasting method and system Download PDFInfo
- Publication number
- CN110414716A CN110414716A CN201910595699.3A CN201910595699A CN110414716A CN 110414716 A CN110414716 A CN 110414716A CN 201910595699 A CN201910595699 A CN 201910595699A CN 110414716 A CN110414716 A CN 110414716A
- Authority
- CN
- China
- Prior art keywords
- lightgbm
- feature
- promise
- model
- enterprise
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Data Mining & Analysis (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Game Theory and Decision Science (AREA)
- Evolutionary Biology (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Educational Administration (AREA)
- Tourism & Hospitality (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Technology Law (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The present invention provides a kind of enterprise based on LightGBM and breaks one's promise probability forecasting method and system, it is analyzed and is understood the method includes the prestige behavior footprint information left in all fields to enterprise, and data are pre-processed, further feature engineering is done in existing data characteristics dimension in combination with business demand, then the dimension of reduction feature is gone using the correlation technique of feature selecting and Feature Dimension Reduction, the machine learning model based on LightGBM is used to remove learning data, the probability risk value that enterprise breaks one's promise and the classification that whether can be broken one's promise are obtained using the model trained.Technical solution of the present invention can further improve financial institution's fraud prevention and reduce the ability of fraction defective, whether realize will appear the accurate identification broken one's promise to enterprise, suitable for solving the problems, such as Corporate finance and credit appraisal, financing risk prevention ability can be effectively improved, can be widely applied to bank to business loan audit and corporate social credit evaluation field.
Description
Technical field
The present invention relates to machine learning techniques field, particularly relates to a kind of enterprise based on LightGBM and break one's promise probabilistic forecasting
Method.
Background technique
Credit is the basis of entire society, and all economic activities, closely bound up with credit in marketing.Currently,
Chinese Enterprises are in the rapid development stage, and influence power is gradually expanded, and have been increasingly becoming the important impetus of socio-economic development.
Therefore the risk management and processing capacity for reinforcing financing market, reduce the financing risk of enterprise, promote the development of financing market, build
It is extremely urgent to found perfect financing risk evaluation and test system;Wherein, Accurate Prediction enterprise breaks one's promise probability, and whether realize can go out enterprise
The accurate identification now broken one's promise is the basis for establishing perfect financing risk evaluation and test system;Whether enterprise breaks one's promise, concerning entire enterprise
Destiny.But does not have also at present and enterprise's probability of breaking one's promise is predicted, realize whether will appear precisely identifying of breaking one's promise to enterprise
Method.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of enterprise based on LightGBM break one's promise probability forecasting method and
System fills up the technological gap of related fields, using the relevant technologies such as big data and artificial intelligence, machine learning, transfers society
The big data modeling innovation enthusiasm of crew, helps various circles of society to solve the problems, such as that Corporate finance provides thinking, further increases
Financing risk prevention ability.
It breaks one's promise probability forecasting method in order to solve the above technical problems, the present invention provides a kind of enterprise based on LightGBM,
The enterprise based on LightGBM probability forecasting method of breaking one's promise includes:
Goodwill behavior footprint message data set is obtained, constructs training dataset, and carry out to the training dataset
Pretreatment and feature extraction construct fisrt feature collection;
It based on the fisrt feature collection, is trained first using LightGBM model, obtains the first LightGBM model;
Then it is trained using tri- models of XGBoost, CatBoost, LightGBM, and extracts each model respectively according to feature weight
Preceding 30 features of the property wanted sequence, construct second feature collection;
It based on the second feature collection, is trained with LightGBM model, obtains the 2nd LightGBM model;
Using the first LightGBM model and the 2nd LightGBM model, according to the prestige behavior foot of enterprise to be predicted
Mark information predicts its probability of breaking one's promise respectively, and to the prediction result of the first the LightGBM model and the 2nd LightGBM model
It is weighted synthesis, obtains final prediction result.
Optionally, the prestige behavior footprint information includes: industrial and commercial equity information, the administrative penalty information, department after desensitization
Method actionable information and civil debt information.
Further, pretreatment and feature extraction are carried out to training dataset, construct fisrt feature collection, comprising:
The training dataset is cleaned, cancelling noise data, and carries out Missing Data Filling;
Feature is done from three statistical nature, cross feature, service feature angles respectively to pretreated training dataset
Engineering carries out characteristic extraction;
Dimensionality reduction is carried out using characteristic of the default feature dimension reduction method to extraction, constructs fisrt feature collection.
Further, the method for Missing Data Filling is carried out to training dataset to fill out for mean value filling, 0 filling, LightGBM
Any one of fill;The method for carrying out dimensionality reduction to the characteristic of extraction is PCA method of descent.
Further, it is based on fisrt feature collection, when being trained using LightGBM model, and is based on second feature collection,
When being trained using LightGBM model, it is all made of cross-validation method and is trained.
Further, the prediction result to the first LightGBM model and the 2nd LightGBM model adds
Power synthesis, obtains final prediction result, specifically:
The prediction result of the first LightGBM model and the 2nd LightGBM model is averaged;By described first
The average value of the prediction result of LightGBM model and the 2nd LightGBM model is as general to breaking one's promise for the enterprise to be predicted
The final prediction result of rate.
Correspondingly, in order to solve the above technical problems, breaking one's promise probability the present invention also provides a kind of enterprise based on LightGBM
Forecasting system, the enterprise based on LightGBM probabilistic forecasting system of breaking one's promise include:
Fisrt feature collection constructs module, for obtaining goodwill behavior footprint message data set, constructs training dataset,
And pretreatment and feature extraction are carried out to the training dataset, construct fisrt feature collection;
First LightGBM model construction module is carried out for being based on the fisrt feature collection using LightGBM model
Training, obtains the first LightGBM model;
Second feature collection construct module, for be based on the fisrt feature collection, using XGBoost, CatBoost,
Tri- models of LightGBM are trained, and extract each model respectively according to preceding 30 features of feature importance ranking, building
Second feature collection;
2nd LightGBM model construction module is instructed for being based on the second feature collection with LightGBM model
Practice, obtains the 2nd LightGBM model;
Fusion Module, for utilizing the first LightGBM model and the 2nd LightGBM model, according to enterprise to be predicted
The prestige behavior footprint information of industry predicts its probability of breaking one's promise respectively, and to the first LightGBM model and the 2nd LightGBM model
Prediction result be weighted synthesis, obtain final prediction result.
Optionally, the prestige behavior footprint information that the fisrt feature collection building module obtains includes: the industry and commerce after desensitization
Equity information, administrative penalty information, persecutio information and civil debt information.
Further, the fisrt feature collection constructs module, is specifically used for:
The training dataset is cleaned, cancelling noise data, and carries out Missing Data Filling;
Feature is done from three statistical nature, cross feature, service feature angles respectively to pretreated training dataset
Engineering carries out characteristic extraction;
Dimensionality reduction is carried out using characteristic of the default feature dimension reduction method to extraction, constructs fisrt feature collection.
Further, the fisrt feature integrates building module to the method for training dataset progress Missing Data Filling as mean value
Filling, 0 filling and LightGBM filling in any one;Characteristic of the fisrt feature collection building module to extraction
It is PCA method of descent according to the method for carrying out dimensionality reduction.
Further, the Fusion Module, is specifically used for:
The prediction result of first LightGBM model and the 2nd LightGBM model is averaged;By the first LightGBM
The average value of the prediction result of model and the 2nd LightGBM model is final as the probability of breaking one's promise to the enterprise to be predicted
Prediction result.
The advantageous effects of the above technical solutions of the present invention are as follows:
The present invention constructs training dataset, and to training data by obtaining goodwill behavior footprint message data set
Collection carries out pretreatment and feature extraction, constructs fisrt feature collection;Based on fisrt feature collection, carried out first using LightGBM model
Training, obtains the first LightGBM model;Then it is trained using tri- models of XGBoost, CatBoost, LightGBM,
And each model is extracted respectively according to preceding 30 features of feature importance ranking, construct second feature collection;Based on second feature collection,
It is trained with LightGBM model, obtains the 2nd LightGBM model;Utilize the first LightGBM model and the 2nd LightGBM
Model predicts its probability of breaking one's promise according to the prestige behavior footprint information of enterprise to be predicted respectively, and to the first LightGBM model
It is weighted synthesis with the prediction result of the 2nd LightGBM model, obtains final prediction result.It breaks one's promise to realize enterprise
The accurate assessment of behavior, improve financial institution's fraud prevention and reduce fraction defective ability, be suitable for solve Corporate finance and
The problem of credit appraisal, can effectively improve financing risk prevention ability, can be widely applied to bank to business loan audit with
And corporate social credit evaluation field.
Detailed description of the invention
Fig. 1 is that the enterprise of the invention based on LightGBM breaks one's promise the schematic illustration of probability forecasting method;
Fig. 2 is that the enterprise of the invention based on LightGBM breaks one's promise the flow diagram of probability forecasting method
Fig. 3 is that the combination of cross feature shows schematic diagram;
The schematic diagram of five folding cross validation of Fig. 4;
Fig. 5 is the feature importance schematic diagram of LightGBM model of the invention.
Specific embodiment
To keep the technical problem to be solved in the present invention, technical solution and advantage clearer, below in conjunction with attached drawing and tool
Body embodiment is described in detail.
First embodiment
As shown in Figure 1, breaking one's promise probability forecasting method the present embodiment provides a kind of enterprise based on LightGBM, it is with enterprise
Center, the prestige behavior footprint information left around enterprise in various aspects, carries out business demand analysis and business demand understands, structure
Training dataset is built, all data are divided, is divided into training set and test set, and training set and test set are carried out respectively
Pretreatment and Feature Engineering;It is designed using the exploitation that training set completes big data algorithm model, is realized using Fusion Model to enterprise
The accurate assessment of industry discreditable behavior.It, which is worth, is to improve using algorithm model to solve the problems, such as that Corporate finance provides new approaches
Financing risk prevention ability.Specifically, the step process of the prediction technique of the present embodiment is as shown in Figure 2, comprising:
S101 obtains goodwill behavior footprint message data set, constructs training dataset, and carry out to training dataset
Pretreatment and feature extraction construct fisrt feature collection;
S102 is based on fisrt feature collection, is trained first with LightGBM model, obtains trained first
LightGBM model;Then it is trained with tri- models of XGBoost, CatBoost, LightGBM, and extracts each model respectively
According to preceding 30 features of feature importance ranking, second feature collection is constructed;
S103 is based on second feature collection, is trained using LightGBM model, obtains trained 2nd LightGBM
Model;
S104, using the first LightGBM model and the 2nd LightGBM model, according to the prestige behavior of enterprise to be predicted
Footprint information predicts its probability of breaking one's promise respectively, and to the prediction result of the first LightGBM model and the 2nd LightGBM model into
Row weighted comprehensive obtains final prediction result.
It should be noted that above-mentioned first LightGBM model and the 2nd LightGBM model are based on different feature sets
And training obtains, to enterprise break one's promise probability predict when, be by the first LightGBM model and the 2nd LightGBM mould
Type is predicted respectively, to obtain two prediction results, and be then in above-mentioned S104 by the first LightGBM model and
The prediction result weighting of 2nd LightGBM model is averaged, and obtains final prediction result.
Prestige behavior footprint information in above-mentioned S101 includes: industrial and commercial equity information, the administrative penalty information, department after desensitization
Method actionable information and civil debt information etc..
It is above-mentioned that pretreatment and feature extraction are carried out to training dataset, construct the process of fisrt feature collection, comprising:
1) training dataset is cleaned, cancelling noise data, and carries out Missing Data Filling;
Wherein, noise data refers to the enterprise only occurred in enterprise's essential information, in people's business trial below
Document, the people's business trial process, Administrative Illegality record, list of owing taxes, pay taxes improper family and limit high consumption list it is any
The enterprise occurred never again in one table, the present embodiment reject such data in preprocessing process.
When carrying out Missing Data Filling, mean value filling, 0 filling and LightGBM filling can choose according to the actual situation
Any one of.
Further, since there are typical imbalanced training sets in the data of this example, therefore in order to avoid to a certain degree
Positive and negative imbalanced training sets phenomenon, the present embodiment use stochastical sampling method, and specific practice is randomly in positive sample and negative sample
Between sample so that positive and negative sample size reaches balanced to a certain extent.
2) Feature Engineering is done to pretreated training dataset, carries out characteristic extraction;
Specifically, the present embodiment does Feature Engineering progress from three statistical nature, cross feature, service feature angles respectively
Characteristic is extracted, therefore the most work of this example is all to be constantly to carry out data cleansing, analyzes business, then not
Feature is looked for disconnectedly.
In cross feature in such as enterprise's essential information registered capital and number of employees can be combined into new feature from
Industry number accounts for average registered capital;People's business try document in each enterprise's status in litigation executed person with it is extensive in case-involving event
The combination performance that the number etc. that case occurs simultaneously is exactly cross feature is executed again.
Whether the status in litigation in service feature such as in people's business trial document there is first trial defendant, was applied
People, executed person, defendant enterprise whether be break one's promise enterprise comparison business performance, owe taxes plant in whether there is place
The enterprise of the tax category such as education surtax, land increment water tax whether be break one's promise enterprise business performance etc.;As shown in Figure 3.
3) dimensionality reduction is carried out using characteristic of the PCA method of descent to extraction, constructs fisrt feature collection.
Specifically, the feature that constructed fisrt feature is concentrated is specifically included that according to feature importance ranking
' the t1_ Date of Incorporation ', ' t1_ industry department code _ ratio_t1_ operation (business) range ', ' t1_ working people
Number ', ' t3_ exact date _ mean', ' t2_ wind up the case the time _ mean', ' t1_ manage (business) range _ ratio_t1_ enterprise (machine
Structure) type ', ' the t2_ status in litigation _ defendant _ case-involving amount of money (member) _ sum', ' t2_ wind up the case the time _ diff_mean', ' t1_ registration
Fund (Wan Yuan) ', ' t2_ winds up the case the time _ min', and ' the case-involving amount of money of t2_ (member) _ sum', ' t2_ winds up the case the time _ max', ' t2_
Count', ' t1_ enterprise (mechanism) type _ ratio_t1_ operation (business) range ', ' onehot_t2_ status in litigation _ be performed
People _ mean', ' t1_ operation (business) range _ ratio_t1_ industry department code ', ' t1_ enterprise (mechanism) type _ ratio_
T1_ industry department code ', ' the case-involving amount of money of t2_ (member) _ mean', ' the t2_ status in litigation _ executed person _ case-involving amount of money (member) _
Sum', ' t3_ exact date _ diff_mean', ' the case-involving amount of money of t2_ (member) _ max', ' t1_ industry department code _ ratio_t1_
Enterprise (mechanism) type ', ' t2_ status in litigation _ appellant _ rat', ' t2_ status in litigation _ quilt _ mean', ' t1_ operation (business)
Range ', ' t3_count', ' onehot_t2_ status in litigation _ defendant _ mean', ' the t2_ status in litigation _ plaintiff _ case-involving amount of money
(member) _ sum', ' t2_ status in litigation _ execution _ mean', ' t1_ industry department code ', ' onehot_t3_ status in litigation _ work as thing
People _ mean', ' t2_ status in litigation _ defendant _ rat', ' t7_ exact date _ max', ' onehot_t3_ status in litigation _ party _
Sum', ' onehot_t3_ status in litigation _ plaintiff _ mean', ' t5_ balance of outstanding taxes (member) _ std', ' t2_ status in litigation _ defendant _
Count', ' onehot_t3_ status in litigation _ applicant _ mean', ' t3_ status in litigation _ quilt _ mean', ' onehot_t2_ lawsuit
Status _ appellant _ mean', ' onehot_t3_ status in litigation _ defendant _ mean', ' t5_ exact date _ diff_mean', '
Onehot_t2_ status in litigation _ plaintiff _ mean', ' t1_ enterprise (mechanism) type ', ' t5_ exact date _ min', ' t5_
Count', ' onehot_t3_ status in litigation _ applicant _ sum', ' t5_ exact date _ mean', ' t6_ identification date _ min', '
Onehot_t3_ status in litigation _ defendant _ sum', ' t5_ balance of outstanding taxes (member) _ max', ' onehot_t2_ status in litigation _ application are held
Pedestrian _ mean', ' t4_ exact date _ min', ' the t2_ status in litigation _ appellant _ case-involving amount of money (member) _ sum', ' onehot_
T5_ owes taxes kind _ value-added tax _ mean', ' onehot_t3_ status in litigation _ plaintiff _ sum', ' t5_ balance of outstanding taxes (member) _
Min', ' onehot_t3_ status in litigation _ nan_mean', ' t7_ exact date _ min', ' t2_ status in litigation _ plaintiff _ rat', '
T2_ status in litigation _ plaintiff _ count', ' the onehot_t3_ status in litigation _ defending party to the application _ sum', ' onehot_t3_ lawsuit
Position _ the defending party to the application _ mean', ' onehot_t2_ status in litigation _ appellee _ mean', ' t5_ owe taxes kind _ income tax _
Sum', ' t6_count', ' t5_ exact date _ max', ' onehot_t2_ status in litigation _ first trial defendant _ mean', ' onehot_
T5_ owes taxes kind _ Tax for maintaining and building cities _ sum', ' onehot_t5_ owe taxes kind _ Tax for maintaining and building cities _ mean', '
Onehot_t5_ owes taxes kind _ stamp tax _ mean', ' the onehot_t2_ status in litigation _ defending party to the application _ mean', ' t7_ specific day
Phase _ mean', ' t7_count', ' onehot_t5_ owe taxes kinds _ 10106 | Individual Income Tax _ mean', ' onehot_t5_ institute
Tax arrear kinds _ 10109 | Tax for maintaining and building cities _ mean', ' t4_count', ' t6_ assert date _ mean', ' t5_ owe taxes plant _
Income tax _ mean', ' onehot_t5_ owe taxes kind _ value-added tax _ sum', ' the t2_ status in litigation _ appellee _ case-involving amount of money
(member) _ sum', ' onehot_t5_ owe taxes kinds _ 10111 | stamp tax _ mean', ' t2_ status in litigation _ application executor _
Rat', ' onehot_t5_ owe taxes kind _ Individual Income Tax _ mean', ' onehot_t3_ status in litigation _ nan_sum', '
Onehot_t5_ owes taxes kind _ Individual Income Tax _ sum', ' onehot_t2_ status in litigation _ appellee (first trial plaintiff) _
Mean', ' onehot_t5_ owe taxes kind _ enterprise income tax _ mean', ' the t2_ status in litigation _ applicant _ case-involving amount of money (member) _
Sum', ' onehot_t3_ status in litigation _ appellee _ mean', ' onehot_t3_ status in litigation _ appellee _ sum', '
Onehot_t3_ status in litigation _ appellant _ mean', ' onehot_t5_ owe taxes kind _ enterprise income tax _ sum', ' onehot_
T3_ status in litigation _ defendant/appellee _ mean', ' onehot_t3_ status in litigation _ applicant: defendant _ mean', '
Onehot_t3_ status in litigation _ applicant: defendant _ sum', ' the t2_ status in litigation _ first trial defendant _ case-involving amount of money (member) _ sum', '
T2_ status in litigation _ application executor _ count', ' onehot_t5_ owe taxes kind _ stamp tax _ sum', ' onehot_t3_ lawsuit
Status _ defendant/defendant/appellee/the defending party to the application _ sum', ' onehot_t5_ owe taxes kinds _ 10111 | stamp tax _
Sum', ' the t2_ status in litigation _ application executor _ case-involving amount of money (member) _ sum', ' onehot_t3_ status in litigation _ appellant's quilt
Announcement _ mean', ' onehot_t2_ status in litigation _ applicant _ mean', ' onehot_t2_ status in litigation _ appellant's (first trial quilt
Accuse) _ mean', ' onehot_t3_ status in litigation _ review applicant _ mean', ' onehot_t5_, which owes taxes, plants _ 10112 | cities and towns
Land use charge _ mean', ' onehot_t5_ owe taxes kinds _ 10113 | increment tax on land value _ sum', ' onehot_t5_ are owed taxes
Kinds _ 10113 | increment tax on land value _ mean', ' onehot_t5_ owe taxes and plant _ 30203 | educational expenses _ sum', ' onehot_
T5_ owes taxes kinds _ 30203 | educational expenses _ mean', ' onehot_t5_ owe taxes kinds _ 30216 | local educational adds _
Sum', ' onehot_t5_ owe taxes kinds _ 30216 | local educational is additional _ mean', ' onehot_t2_ status in litigation _ appealed
People (first trial plaintiff, countercharge defendant) _ mean', ' onehot_t2_ status in litigation _ request reconsideration people _ mean', ' onehot_t2_
Status in litigation _ the third party _ mean', ' onehot_t2_ status in litigation _ appellee (first trial defendant) _ mean', ' onehot_
T2_ status in litigation _ application reviews people _ mean', ' onehot_t5_ owe taxes kind _ enterprise income tax value-added tax _ sum', '
Onehot_t5_ owes taxes kind _ enterprise income tax value-added tax _ mean', ' onehot_t2_ status in litigation _ outsider _ mean', '
Onehot_t5_ owes taxes kind _ increment tax on land value _ sum', ' onehot_t5_ owe taxes kind _ increment tax on land value _ mean', '
Onehot_t5_ owe taxes kind _ local educational it is additional _ sum', ' onehot_t5_ owe taxes kinds _ 10112 | urban land use
Tax _ sum', ' onehot_t2_ status in litigation _ appellee (first trial defendant, countercharge plaintiff) _ mean', ' onehot_t2_ lawsuit
Status _ opponent (executed person) _ mean', ' onehot_t2_ status in litigation _ defendant _ mean', ' onehot_t2_ lawsuit
Status _ be applied executor _ mean', ' the onehot_t2_ status in litigation _ defending party to the application (first trial defendant) _ mean', ' onehot_
T2_ status in litigation _ the defending party to the application (first sentence defendant, second trial appellee) _ mean', ' onehot_t2_ status in litigation _ be applied
People (first sentence plaintiff, second trial appellant) _ mean', ' the onehot_t2_ status in litigation _ defending party to the application (first sentence plaintiff) _ mean', '
Onehot_t2_ status in litigation _ by complainant _ mean', ' onehot_t2_ status in litigation _ by careful applicant _ mean', '
Onehot_t2_ status in litigation _ the defendant work unit _ mean', ' onehot_t5_ owe taxes kinds _ 10103 | Sales Tax _ sum', '
Onehot_t5_ owes taxes kinds _ 10110 | house property tax _ mean', ' onehot_t5_ owe taxes and plant _ 10103 | Sales Tax _
Mean', ' onehot_t5_ owe taxes kinds _ 10104 | enterprise income tax _ sum', ' onehot_t5_ owe taxes kinds _ 10104 | it looks forward to
Industry income tax _ mean', ' onehot_t5_ owe taxes kinds _ 10106 | Individual Income Tax _ sum', ' onehot_t2_ status in litigation _
Defendant (countercharge plaintiff) _ mean', ' onehot_t5_ owe taxes and plant _ 10109 | Tax for maintaining and building cities _ sum', ' onehot_
T2_ status in litigation _ defendant (plaintiff) _ mean', ' onehot_t5_ owe taxes kinds _ 10110 | house property tax _ sum', ' onehot_
T5_ owe taxes kind _ local educational it is additional _ mean', ' onehot_t2_ status in litigation _ opponent _ mean', ' onehot_t2_ tells
Status _ compensation claimant _ mean' is disputed, ' onehot_t2_ status in litigation _ review applicant (first sentence plaintiff, second trial appellant) _
mean'。
Specifically, the feature importance for the LightGBM model performance that the present embodiment is selected is as shown in Figure 5.
Further, in order to improve the generalization ability of model, this example uses 5 foldings intersection when carrying out model training and tests
Demonstration has trained 5 LightGBM models, then takes mean value as last pre- the prediction result of 5 LightGBM models
It surveys as a result, as shown in Figure 4.It is using the advantage of cross-validation method, over-fitting can be effectively prevented, enhances the extensive energy of model
Power.
In addition, Auc reaction model is strong and weak to positive negative sample sequencing ability, size and precision to score are not required.
The present embodiment carries out the fusion of model using the method for ranking, it can be merged between multiple models using ranking faster
Difference, without Weighted Fusion probability;The formula of Ranking is as follows:
Specifically, the process that the present embodiment carries out parameter adjustment to LightGBM model is as follows:
It selects higher learning rate first, is to accelerate convergent speed in this way near general 0.1.Then to decision
The basic parameter tune of tree is joined, followed by the tune ginseng of regularization parameter, finally reduces learning rate, is finally accurate in order to improve here
Rate;Specific practice is:
Step1. the setting of learning rate and the number of iterations: learning_rate=0.1, the number of iterations n_estimator are first
If then a biggish number checks that optimal the number of iterations, this example are provided that 8000 in the result of cross validation.
Step2. the depth capacity and leaf node number set: being the most important parameters for improving accuracy, max_depth here
=5, num_leaves=2^ (max_depth), but its value setting should be less than 2^ (max_depth), otherwise will lead to
Over-fitting.This step can carry out tuning, first coarse adjustment fine tuning again to the two parameters simultaneously.
Step3.Min_data_in_leaf and min_sum_hessian_in_leaf: this step is quasi- in order to reduce
It closes, min_data_in_leaf is a critically important parameter, is also min_child_samples, its value depends on training
The number of samples and num_leaves. of data be arranged it is larger can to avoid generate a too deep tree, it is likely that leading
Cause poor fitting.Min_sum_hessian_in_leaf: being also min_child_weight, is the minimum of a node split
The sum of Hessian value.
The parameter value that this example is arranged after being adjusted is 20,0.001 respectively.
Step4.feature_fraction and bagging_fraction: the two parameters are provided to reduce over-fitting
's.Bagging_feaction=1.0, feature_fraction=0.7.
Step5. there is no adjust for the default value that this example of regularization parameter directly uses.
Step6. it reduces learning rate: learning rate being scheduled on 0.005 according to the performance of model.
Second embodiment
It breaks one's promise probabilistic forecasting system the present embodiment provides a kind of enterprise based on LightGBM, it should enterprise based on LightGBM
Industry probabilistic forecasting system of breaking one's promise includes:
Fisrt feature collection constructs module, for obtaining goodwill behavior footprint message data set, constructs training dataset,
And pretreatment and feature extraction are carried out to training dataset, construct fisrt feature collection;
First LightGBM model construction module, for being trained using LightGBM model based on fisrt feature collection,
Obtain the first LightGBM model;
Second feature collection constructs module, for being based on fisrt feature collection, uses XGBoost, CatBoost, LightGBM tri-
A model is trained, and extracts each model respectively according to preceding 30 features of feature importance ranking, constructs second feature collection;
2nd LightGBM model construction module is trained with LightGBM model, is obtained for being based on second feature collection
2nd LightGBM model;
Fusion Module, for utilizing the first LightGBM model and the 2nd LightGBM model, according to enterprise to be predicted
Prestige behavior footprint information predicts its probability of breaking one's promise respectively, and to the pre- of the first LightGBM model and the 2nd LightGBM model
It surveys result and is weighted synthesis, obtain final prediction result.
The enterprise based on LightGBM of the present embodiment break one's promise in probabilistic forecasting system and above-mentioned first embodiment based on
The enterprise of LightGBM breaks one's promise probability forecasting method reciprocal correspondence, wherein each modular unit is realized in the system function with
Each process step in the above method corresponds;Therefore details are not described herein.
The present invention constructs training dataset, and to training data by obtaining goodwill behavior footprint message data set
Collection carries out pretreatment and feature extraction, constructs fisrt feature collection;Based on fisrt feature collection, carried out first using LightGBM model
Training, obtains the first LightGBM model;Then it is trained using tri- models of XGBoost, CatBoost, LightGBM,
And each model is extracted respectively according to preceding 30 features of feature importance ranking, construct second feature collection;Based on second feature collection,
It is trained with LightGBM model, obtains the 2nd LightGBM model;Utilize the first LightGBM model and the 2nd LightGBM
Model predicts its probability of breaking one's promise according to the prestige behavior footprint information of enterprise to be predicted respectively, and to the first LightGBM model
It is weighted synthesis with the prediction result of the 2nd LightGBM model, obtains final prediction result.It breaks one's promise to realize enterprise
The accurate assessment of behavior, improve financial institution's fraud prevention and reduce fraction defective ability, be suitable for solve Corporate finance and
The problem of credit appraisal, can effectively improve financing risk prevention ability, can be widely applied to bank to business loan audit with
And corporate social credit evaluation field.
In addition, it should be noted that, it should be understood by those skilled in the art that, the embodiment of the embodiment of the present invention can provide
For method, apparatus or computer program product.Therefore, it is real that complete hardware embodiment, complete software can be used in the embodiment of the present invention
Apply the form of example or embodiment combining software and hardware aspects.Moreover, the embodiment of the present invention can be used it is one or more its
In include computer usable program code computer-usable storage medium (including but not limited to magnetic disk storage, CD-ROM,
Optical memory etc.) on the form of computer program product implemented.
The embodiment of the present invention be referring to according to the method for the embodiment of the present invention, terminal device (system) and computer program
The flowchart and/or the block diagram of product describes.It should be understood that flowchart and/or the block diagram can be realized by computer program instructions
In each flow and/or block and flowchart and/or the block diagram in process and/or box combination.It can provide these
Computer program instructions to general purpose computer, Embedded Processor or other programmable data processing terminal devices processor with
A machine is generated, so that generating by the instruction that computer or the processor of other programmable data processing terminal devices execute
For realizing the function of being specified in one or more flows of the flowchart and/or one or more blocks of the block diagram
Device.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing terminal devices
In computer-readable memory operate in a specific manner, so that instruction stored in the computer readable memory generates packet
The manufacture of command device is included, which realizes in one side of one or more flows of the flowchart and/or block diagram
The function of being specified in frame or multiple boxes.These computer program instructions can also be loaded at computer or other programmable datas
It manages on terminal device, so that executing series of operation steps on computer or other programmable terminal equipments to generate computer
The processing of realization, so that the instruction executed on computer or other programmable terminal equipments is provided for realizing in flow chart one
The step of function of being specified in a process or multiple processes and/or one or more blocks of the block diagram.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications can be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of range of embodiment of the invention.
It should also be noted that, herein, the terms "include", "comprise" or its any other variant are intended to non-
It is exclusive to include, so that process, method, article or terminal device including a series of elements are not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or terminal
The intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence "including a ...", is not arranged
Except there is also other identical elements in process, method, article or the terminal device for including the element.
The above is a preferred embodiment of the present invention, it is noted that for those skilled in the art
For, without departing from the principles of the present invention, several improvements and modifications can also be made, these improvements and modifications
It should be regarded as protection scope of the present invention.
Claims (10)
- The probability forecasting method 1. a kind of enterprise based on LightGBM breaks one's promise characterized by comprisingGoodwill behavior footprint message data set is obtained, constructs training dataset, and located in advance to the training dataset Reason and feature extraction construct fisrt feature collection;It based on the fisrt feature collection, is trained first using LightGBM model, obtains the first LightGBM model;Then It is trained using tri- models of XGBoost, CatBoost, LightGBM, and extracts each model respectively according to feature importance Preceding 30 features of sequence construct second feature collection;It based on the second feature collection, is trained with LightGBM model, obtains the 2nd LightGBM model;Using the first LightGBM model and the 2nd LightGBM model, believed according to the prestige behavior footprint of enterprise to be predicted Breath predicts its probability of breaking one's promise respectively, and carries out to the prediction result of the first LightGBM model and the 2nd LightGBM model Weighted comprehensive obtains final prediction result.
- 2. the enterprise based on LightGBM breaks one's promise probability forecasting method as described in claim 1, which is characterized in that the prestige Behavior footprint information includes: industrial and commercial equity information, administrative penalty information, persecutio information and the civil debt after desensitization Information.
- 3. the enterprise based on LightGBM breaks one's promise probability forecasting method as described in claim 1, which is characterized in that described to institute It states training dataset and carries out pretreatment and feature extraction, construct fisrt feature collection, comprising:The training dataset is cleaned, cancelling noise data, and carries out Missing Data Filling;Feature Engineering is done from three statistical nature, cross feature, service feature angles respectively to pretreated training dataset, Carry out characteristic extraction;Dimensionality reduction is carried out using characteristic of the default feature dimension reduction method to extraction, constructs fisrt feature collection.
- 4. the enterprise based on LightGBM breaks one's promise probability forecasting method as claimed in claim 3, which is characterized in that training number According to the method for integrating progress Missing Data Filling as any one in mean value filling, 0 filling and LightGBM filling;To extraction Characteristic carry out dimensionality reduction method be PCA method of descent.
- 5. the enterprise based on LightGBM breaks one's promise probability forecasting method as described in claim 1, which is characterized in that based on described Fisrt feature collection, when being trained using LightGBM model, and be based on the second feature collection, using LightGBM model into When row training, it is all made of cross-validation method and is trained.
- The probability forecasting method 6. enterprise as described in any one in claim 1-5 based on LightGBM breaks one's promise, which is characterized in that The prediction result to the first LightGBM model and the 2nd LightGBM model is weighted synthesis, obtains final pre- Survey as a result, specifically:The prediction result of the first LightGBM model and the 2nd LightGBM model is averaged;By described first The average value of the prediction result of LightGBM model and the 2nd LightGBM model is as general to breaking one's promise for the enterprise to be predicted The final prediction result of rate.
- The probabilistic forecasting system 7. a kind of enterprise based on LightGBM breaks one's promise characterized by comprisingFisrt feature collection constructs module, for obtaining goodwill behavior footprint message data set, constructs training dataset, and right The training dataset carries out pretreatment and feature extraction, constructs fisrt feature collection;First LightGBM model construction module, for being trained using LightGBM model based on the fisrt feature collection, Obtain the first LightGBM model;Second feature collection constructs module, for being based on the fisrt feature collection, uses XGBoost, CatBoost, LightGBM tri- A model is trained, and extracts each model respectively according to preceding 30 features of feature importance ranking, constructs second feature collection;2nd LightGBM model construction module is trained with LightGBM model, is obtained for being based on the second feature collection 2nd LightGBM model;Fusion Module, for utilizing the first LightGBM model and the 2nd LightGBM model, according to enterprise to be predicted Prestige behavior footprint information predicts its probability of breaking one's promise respectively, and to the pre- of the first LightGBM model and the 2nd LightGBM model It surveys result and is weighted synthesis, obtain final prediction result.
- 8. the enterprise based on LightGBM breaks one's promise probabilistic forecasting system as claimed in claim 7, which is characterized in that described first Feature set constructs module, is specifically used for:The training dataset is cleaned, cancelling noise data, and carries out Missing Data Filling;Feature Engineering is done from three statistical nature, cross feature, service feature angles respectively to pretreated training dataset, Carry out characteristic extraction;Dimensionality reduction is carried out using characteristic of the default feature dimension reduction method to extraction, constructs fisrt feature collection.
- 9. the enterprise based on LightGBM breaks one's promise probabilistic forecasting system as claimed in claim 8, which is characterized in that described first It is that mean value filling, 0 filling and LightGBM are filled out that feature set, which constructs the method that module carries out Missing Data Filling to training dataset, Any one in filling;The fisrt feature integrates building module and carries out the method for dimensionality reduction to the characteristic of extraction as PCA dimensionality reduction Method.
- The probabilistic forecasting system 10. such as described in any item enterprises based on LightGBM of claim 7-9 break one's promise, feature exist In the Fusion Module is specifically used for:The prediction result of first LightGBM model and the 2nd LightGBM model is averaged;By the first LightGBM model Final prediction with the average value of the prediction result of the 2nd LightGBM model as the probability of breaking one's promise to the enterprise to be predicted As a result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910595699.3A CN110414716B (en) | 2019-07-03 | 2019-07-03 | LightGBM-based enterprise confidence loss probability prediction method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910595699.3A CN110414716B (en) | 2019-07-03 | 2019-07-03 | LightGBM-based enterprise confidence loss probability prediction method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110414716A true CN110414716A (en) | 2019-11-05 |
CN110414716B CN110414716B (en) | 2022-05-20 |
Family
ID=68358786
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910595699.3A Active CN110414716B (en) | 2019-07-03 | 2019-07-03 | LightGBM-based enterprise confidence loss probability prediction method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110414716B (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110941963A (en) * | 2019-11-29 | 2020-03-31 | 福州大学 | Text attribute viewpoint abstract generation method and system based on sentence emotion attributes |
CN111242364A (en) * | 2020-01-07 | 2020-06-05 | 上海钧正网络科技有限公司 | Neural network-based vehicle fault and comfort prediction method, device, terminal and medium |
CN111291798A (en) * | 2020-01-21 | 2020-06-16 | 北京工商大学 | User basic attribute prediction method based on ensemble learning |
CN111310785A (en) * | 2020-01-15 | 2020-06-19 | 杭州华网信息技术有限公司 | National power grid mechanical external damage prediction method |
CN111798303A (en) * | 2020-07-06 | 2020-10-20 | 浙江公共安全技术研究院有限公司 | Method for assessing fulfillment ability of court executives |
CN112215616A (en) * | 2020-11-30 | 2021-01-12 | 四川新网银行股份有限公司 | Method and system for automatically identifying abnormal fund transaction based on network |
CN112418010A (en) * | 2020-11-09 | 2021-02-26 | 电子科技大学 | Equipment temperature monitoring method based on LightGBM fusion model |
CN113205271A (en) * | 2021-05-12 | 2021-08-03 | 国家税务总局山东省税务局 | Method for evaluating enterprise income tax risk based on machine learning |
CN113240509A (en) * | 2021-05-18 | 2021-08-10 | 重庆邮电大学 | Loan risk assessment method based on multi-source data federal learning |
CN113537576A (en) * | 2021-06-25 | 2021-10-22 | 合肥工业大学 | Method and system for predicting financial predicament of listed enterprises |
CN114418024A (en) * | 2022-01-26 | 2022-04-29 | 深圳前海微众银行股份有限公司 | Academic calendar prediction method, academic calendar prediction system, terminal device and storage medium |
CN115409104A (en) * | 2022-08-25 | 2022-11-29 | 贝壳找房(北京)科技有限公司 | Method, apparatus, device, medium and program product for identifying object type |
CN114418024B (en) * | 2022-01-26 | 2024-05-31 | 深圳前海微众银行股份有限公司 | Method, system, terminal device and storage medium for academic prediction |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160048766A1 (en) * | 2014-08-13 | 2016-02-18 | Vitae Analytics, Inc. | Method and system for generating and aggregating models based on disparate data from insurance, financial services, and public industries |
KR101599675B1 (en) * | 2015-07-02 | 2016-03-03 | 한국기업데이터 주식회사 | Apparatus and method for predicting degree of corporation credit risk using corporation news searching technology based on big data technology |
CN105868847A (en) * | 2016-03-24 | 2016-08-17 | 车智互联(北京)科技有限公司 | Shopping behavior prediction method and device |
CN108985489A (en) * | 2018-06-08 | 2018-12-11 | 阿里巴巴集团控股有限公司 | A kind of Risk Forecast Method, risk profile device and terminal device |
CN109034658A (en) * | 2018-08-22 | 2018-12-18 | 重庆邮电大学 | A kind of promise breaking consumer's risk prediction technique based on big data finance |
CN109063931A (en) * | 2018-09-06 | 2018-12-21 | 盈盈(杭州)网络技术有限公司 | A kind of model method for predicting freight logistics driver Default Probability |
CN109255651A (en) * | 2018-08-22 | 2019-01-22 | 重庆邮电大学 | A kind of search advertisements conversion intelligent Forecasting based on big data |
CN109389247A (en) * | 2018-09-27 | 2019-02-26 | 智庭(北京)智能科技有限公司 | A kind of region house rent prediction technique based on big data |
CN109509033A (en) * | 2018-12-14 | 2019-03-22 | 重庆邮电大学 | A kind of user buying behavior big data prediction technique under consumer finance scene |
CN109740764A (en) * | 2019-01-10 | 2019-05-10 | 博拉网络股份有限公司 | A kind of small business based on big data breaks one's promise prediction technique |
CN109784979A (en) * | 2018-12-19 | 2019-05-21 | 重庆邮电大学 | A kind of supply chain needing forecasting method of big data driving |
CN109961248A (en) * | 2017-12-25 | 2019-07-02 | 顺丰科技有限公司 | Waybill complains prediction technique, device, equipment and its storage medium |
-
2019
- 2019-07-03 CN CN201910595699.3A patent/CN110414716B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160048766A1 (en) * | 2014-08-13 | 2016-02-18 | Vitae Analytics, Inc. | Method and system for generating and aggregating models based on disparate data from insurance, financial services, and public industries |
KR101599675B1 (en) * | 2015-07-02 | 2016-03-03 | 한국기업데이터 주식회사 | Apparatus and method for predicting degree of corporation credit risk using corporation news searching technology based on big data technology |
CN105868847A (en) * | 2016-03-24 | 2016-08-17 | 车智互联(北京)科技有限公司 | Shopping behavior prediction method and device |
CN109961248A (en) * | 2017-12-25 | 2019-07-02 | 顺丰科技有限公司 | Waybill complains prediction technique, device, equipment and its storage medium |
CN108985489A (en) * | 2018-06-08 | 2018-12-11 | 阿里巴巴集团控股有限公司 | A kind of Risk Forecast Method, risk profile device and terminal device |
CN109034658A (en) * | 2018-08-22 | 2018-12-18 | 重庆邮电大学 | A kind of promise breaking consumer's risk prediction technique based on big data finance |
CN109255651A (en) * | 2018-08-22 | 2019-01-22 | 重庆邮电大学 | A kind of search advertisements conversion intelligent Forecasting based on big data |
CN109063931A (en) * | 2018-09-06 | 2018-12-21 | 盈盈(杭州)网络技术有限公司 | A kind of model method for predicting freight logistics driver Default Probability |
CN109389247A (en) * | 2018-09-27 | 2019-02-26 | 智庭(北京)智能科技有限公司 | A kind of region house rent prediction technique based on big data |
CN109509033A (en) * | 2018-12-14 | 2019-03-22 | 重庆邮电大学 | A kind of user buying behavior big data prediction technique under consumer finance scene |
CN109784979A (en) * | 2018-12-19 | 2019-05-21 | 重庆邮电大学 | A kind of supply chain needing forecasting method of big data driving |
CN109740764A (en) * | 2019-01-10 | 2019-05-10 | 博拉网络股份有限公司 | A kind of small business based on big data breaks one's promise prediction technique |
Non-Patent Citations (1)
Title |
---|
邱耀等: "基于XGBoost算法的用户行为预测与风险分析", 《工业控制计算机》 * |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110941963A (en) * | 2019-11-29 | 2020-03-31 | 福州大学 | Text attribute viewpoint abstract generation method and system based on sentence emotion attributes |
CN111242364A (en) * | 2020-01-07 | 2020-06-05 | 上海钧正网络科技有限公司 | Neural network-based vehicle fault and comfort prediction method, device, terminal and medium |
CN111310785A (en) * | 2020-01-15 | 2020-06-19 | 杭州华网信息技术有限公司 | National power grid mechanical external damage prediction method |
CN111291798A (en) * | 2020-01-21 | 2020-06-16 | 北京工商大学 | User basic attribute prediction method based on ensemble learning |
CN111798303A (en) * | 2020-07-06 | 2020-10-20 | 浙江公共安全技术研究院有限公司 | Method for assessing fulfillment ability of court executives |
CN112418010A (en) * | 2020-11-09 | 2021-02-26 | 电子科技大学 | Equipment temperature monitoring method based on LightGBM fusion model |
CN112215616A (en) * | 2020-11-30 | 2021-01-12 | 四川新网银行股份有限公司 | Method and system for automatically identifying abnormal fund transaction based on network |
CN112215616B (en) * | 2020-11-30 | 2021-04-30 | 四川新网银行股份有限公司 | Method and system for automatically identifying abnormal fund transaction based on network |
CN113205271A (en) * | 2021-05-12 | 2021-08-03 | 国家税务总局山东省税务局 | Method for evaluating enterprise income tax risk based on machine learning |
CN113240509A (en) * | 2021-05-18 | 2021-08-10 | 重庆邮电大学 | Loan risk assessment method based on multi-source data federal learning |
CN113240509B (en) * | 2021-05-18 | 2022-04-22 | 重庆邮电大学 | Loan risk assessment method based on multi-source data federal learning |
CN113537576A (en) * | 2021-06-25 | 2021-10-22 | 合肥工业大学 | Method and system for predicting financial predicament of listed enterprises |
CN114418024A (en) * | 2022-01-26 | 2022-04-29 | 深圳前海微众银行股份有限公司 | Academic calendar prediction method, academic calendar prediction system, terminal device and storage medium |
CN114418024B (en) * | 2022-01-26 | 2024-05-31 | 深圳前海微众银行股份有限公司 | Method, system, terminal device and storage medium for academic prediction |
CN115409104A (en) * | 2022-08-25 | 2022-11-29 | 贝壳找房(北京)科技有限公司 | Method, apparatus, device, medium and program product for identifying object type |
Also Published As
Publication number | Publication date |
---|---|
CN110414716B (en) | 2022-05-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110414716A (en) | A kind of enterprise based on LightGBM breaks one's promise probability forecasting method and system | |
Du et al. | CUS-heterogeneous ensemble-based financial distress prediction for imbalanced dataset with ensemble feature selection | |
Zhang | Financial viability analysis and capital structure optimization in privatized public infrastructure projects | |
Mostafavi et al. | Ex-ante policy analysis in civil infrastructure systems | |
Arciero et al. | Exploring agent-based methods for the analysis of payment systems: A crisis model for StarLogo TNG | |
Robinson | A richness index for baselining climate change adaptations in small island developing states | |
Clarke | The long road to reformulating the understanding of directors' duties: Legalizing team production theory | |
Arora et al. | PREDICTION & WARNING: a method to improve student's performance | |
Akinwale et al. | Economic evaluation of Nigerian marginal oil and gas field using financial simulation analysis | |
Biswas et al. | Automated credit assessment framework using ETL process and machine learning | |
Cung | Gross domestic product and foreign direct investment: Empirical evidence from Vietnam | |
Amin et al. | External debt, investment and economic growth: Evidence from Nigeria | |
CN109523185A (en) | The method, apparatus and storage medium of collection scorecard are generated based on artificial intelligence | |
Sykes et al. | Using scenarios to investigate the long-term future of copper mining and guide exploration targeting strategies | |
CN109711848A (en) | A kind of matching system and its construction method, matching process of financial transaction | |
Ayadi et al. | Construction of financial social accounting matrix for Tunisia | |
Vydhyam et al. | A study on green banking initiatives of selected private and public sector banks in chittoor (dt.), andhra pradesh | |
Shi et al. | An optimal decision assessment model based on the acceptable maximum LGD of commercial banks and its application | |
Bassi | Modeling US energy with threshold 21 (T21) | |
Theuri et al. | The impact of Artficial Intelligence and how it is shaping banking | |
Göndör et al. | Fiscal policy, the main tool to influence the capital markets’ strength | |
Guo et al. | Research and verification analysis on early warning model through support vector machine algorithm | |
Thamprasert et al. | Simulated trial and error experiments on productivity | |
Iannuzzi et al. | Sustainability and the circular economy: The role of the financial sector and the response of Intesa Sanpaolo | |
Ogot | An Investigation of Political, Economic and Environmental Risk Factors That Affect Successful Implementation of Public Private Partnership Projects In The Kenyan Road Sub-Sector |
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 |