Summary of the invention
Based on this, it is necessary in view of the above technical problems, provide a kind of wind for improving air control model accuracy rate and stabilityControl model modeling, business risk appraisal procedure, device, computer equipment and storage medium.
A kind of air control model modelling approach, comprising:
Obtain positive sample data set, negative sample data set and validation data set；
The validation data set, the positive sample data set and the negative sample data set are inputted into engineering to be trainedIt practises model to be trained, until obtaining enterprise's Rating Model after meeting training condition；
Using the validation data set, the positive sample data set and the negative sample data set as abnormal test scoring mouldThe sample characteristics of type obtain abnormal test Rating Model；
By enterprise's Rating Model in conjunction with abnormal test Rating Model, air control model is obtained.
In one of the embodiments, by the validation data set, the positive sample data set and the negative sample dataCollection inputs machine learning model to be trained and is trained, until after meeting training condition, the step of obtaining enterprise's Rating Model,Include:
The positive sample data set and the negative sample data set are inputted into machine learning model to be trained；
The machine learning model to be trained is based on the positive sample data set and the negative sample data set, according to phaseThe data characteristics vector answered, the machine learning model after being trained；
The machine learning model that the validation data set inputs after the training is scored, business risk is obtained and commentsPoint；
When business risk scoring within a preset range, meet training condition, obtain enterprise's Rating Model.
In one of the embodiments, by the validation data set, the positive sample data set and the negative sample dataThe step of collecting the sample characteristics as abnormal test Rating Model, obtaining abnormal test Rating Model, comprising:
Using the validation data set, the positive sample data set and the negative sample data set as abnormal test scoring mouldThe sample characteristics of type carry out feature extraction, obtain the feature vector of each sample characteristics；
Each described eigenvector is clustered, each group variety is obtained；
It is distributed according to the feature space of each group variety, determines to contribute maximum sample characteristics in each group variety；
Using the maximum sample characteristics of contribution degree as the assessment feature of air control model, abnormal test Rating Model is obtained.
In one of the embodiments, by enterprise's Rating Model in conjunction with abnormal test Rating Model, air control is obtainedThe step of model, comprising:
Based on the first business risk scoring of enterprise's Rating Model output and the of the output of abnormal test Rating ModelThe mode that the scoring of two business risks is averaged is combined, and obtains air control model.
The step of obtaining positive sample data set and negative sample data set, validation data set in one of the embodiments, is wrappedIt includes:
Obtain the enterprise's financial data sample, invoice information sample, enterprise operation data sample of each enterprise；
The enterprise's financial data sample, the invoice information sample, the enterprise operation data sample are based on dataLabel is matched with positive negative sample inventory, obtains positive sample data set and negative sample data set；
By the financial data sample, the invoice information sample, the enterprise operation data sample of non-successful matchAs validation data set.
A kind of business risk appraisal procedure, which comprises
Obtain enterprise's financial data, the invoice information, enterprise operation data of enterprise to be scored；
By the enterprise's financial data, the invoice information, the enterprise operation data input air control model intoThe assessment of row business risk, obtains business risk assessment result.
A kind of air control model modeling device includes: by described device
Data set acquisition module, for obtaining positive sample data set and negative sample data set, validation data set；
Enterprise's Rating Model training module is used for the validation data set, the positive sample data set and the negative sampleNotebook data collection inputs machine learning model to be trained and is trained, until obtaining enterprise's Rating Model after meeting training condition；
Abnormal test Rating Model establishes module, for by the validation data set, the positive sample data set and describedSample characteristics of the negative sample data set as abnormal test Rating Model obtain abnormal test Rating Model；
Models coupling module, for enterprise's Rating Model in conjunction with abnormal test Rating Model, to be obtained air control mouldType.
A kind of business risk assessment device, described device include:
Data acquisition module, for obtaining enterprise's financial data, the invoice information, enterprise operation data of enterprise to be scored；
Business risk grading module is used for the enterprise's financial data, the invoice information, the enterprise operation dataThe input air control model carries out business risk scoring, obtains business risk value.
A kind of computer equipment, including memory and processor, the memory are stored with computer program, the processingThe step of device realizes the method when executing the computer program.
A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processorThe step of method is realized when row.
Above-mentioned air control model modelling approach, device, computer equipment and storage medium obtain positive sample data set and negative sampleValidation data set, positive sample data set and negative sample data set are inputted engineering to be trained by notebook data collection, validation data setIt practises model to be trained, until enterprise's Rating Model is obtained, by validation data set, positive sample data set after meeting training conditionSample characteristics with negative sample data set as abnormal test Rating Model obtain abnormal test Rating Model；Enterprise is scoredModel obtains air control model in conjunction with abnormal test Rating Model, and enterprise's Rating Model combination abnormal test Rating Model carries outBusiness risk scoring reduces resultant error, further improves air control model accuracy rate and stability.
Above-mentioned business risk appraisal procedure, device, computer equipment and storage medium, by the enterprise for obtaining enterprise to be scoredIndustry financial data, invoice information, enterprise operation data；The wind that enterprise's financial data, invoice information, enterprise operation data are inputtedControl model carry out business risk assessment, obtain business risk assessment result, air control model by enterprise's financial data, invoice information,Enterprise operation data carry out business risk scoring by enterprise's Rating Model in conjunction with abnormal test Rating Model reduces result mistakeDifference improves the accuracy rate of business risk assessment.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understoodThe application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, notFor limiting the application.
Air control model modelling approach provided by the present application, can be applied in application environment as shown in Figure 1.Wherein, eventuallyEnd 102 is communicated with server 104 by network by network.User passes through 102 pairs of execution air control model modeling sides of terminalThe server 104 of method is configured, so that server 104 obtains positive sample data set, negative sample data set and validation data set；Validation data set, positive sample data set and negative sample data set are inputted machine learning model to be trained to be trained, untilAfter meeting training condition, enterprise's Rating Model is obtained；Using validation data set, positive sample data set and negative sample data set as differentThe sample characteristics for often examining Rating Model, obtain abnormal test Rating Model；By enterprise's Rating Model and abnormal test scoring mouldType combines, and obtains air control model.Wherein, terminal 102 can be, but not limited to be various personal computers, laptop, intelligenceMobile phone, tablet computer and portable wearable device, server 104 can use independent server either multiple server groupsAt server cluster realize.
In one embodiment, as shown in Fig. 2, providing a kind of air control model modelling approach, it is applied to Fig. 1 in this wayIn server for be illustrated, including step S220 to step S280:
Step S220 obtains positive sample data set, negative sample data set and validation data set.
Wherein, positive sample data set refers to by not having risky enterprise or risk lower than the enterprise in the enterprise of preset valueThe data set of the generations such as financial data, invoice information, enterprise operation data, the data in positive sample data have positive sample label；Negative sample data set refers to the enterprise's financial data being higher than in the enterprise of preset value by risk, invoice information, enterprise operation numberAccording to the data set of equal generations, the data in negative sample data set have negative sample label, and validation data set refers to not having riskyEnterprise, risk are higher than the enterprise of preset value or risk lower than the enterprise's financial data in the enterprise of preset value, invoice information, enterpriseThe data set of the generations such as management data, the data that verify data is concentrated do not have positive and negative sample label, the data that verify data is concentratedIt may include the positive sample data set of no positive and negative sample label, the data in negative sample data set.
Validation data set, positive sample data set and negative sample data set are inputted machine learning to be trained by step S240Model is trained, until obtaining enterprise's Rating Model after meeting training condition.
Wherein, validation data set, positive sample data set and negative sample data set are inputted into machine learning model to be trainedIt is trained, so that machine learning model to be trained is trained, machine learning model to be trained obtains each positive sample numberAccording to the feature vector of collection and the feature vector of negative sample data set, the feature vector and negative sample number of positive sample data set are arrived in studyPositive sample is exported according to the feature vector of collection, and according to the feature vector of positive sample data set and the feature vector of negative sample data setThe corresponding business risk scoring of each data in data set, negative sample data set, the machine learning model after being trained；Pass throughValidation data set verifies the machine learning model after training, the corresponding enterprise's wind of each data that output verify data is concentratedDanger scoring after business risk scoring meets training condition, obtains enterprise's Rating Model, and each data that verify data is concentrated are correspondingWhen business risk is scored within a preset range, meet training condition, preset range is concentrated for user according to each verify data eachThe practical risk situation for the enterprise that data indicate determines.
Step S260, using validation data set, positive sample data set and negative sample data set as abnormal test Rating ModelSample characteristics, obtain abnormal test Rating Model.
Wherein, the data in validation data set, positive sample data set and negative sample data set are corresponding as every profession and tradeSample characteristics cluster the sample characteristics of every profession and trade, obtain each group variety, are distributed according to the feature space of each group variety, determineMaximum sample characteristics are contributed in each group variety out, the sample characteristics in each group variety are subjected to dimensionality reduction, obtain feature space distribution；It is logicalIt crosses and is distributed according to the feature space of each group variety, determine to contribute maximum sample characteristics in each group variety；By the maximum sample of contribution degreeAssessment feature of the eigen as air control model obtains abnormal test Rating Model.According to company information to the enterprise of different industriesIndustry is divided, and for each different industry, the sample characteristics in the sector is taken to cluster, and cluster is that (sample refers to for researchMark) classification problem a kind of statistical analysis technique, while be also data mining an important algorithm；The mode of cluster can beIt is several below: k-means algorithm, K-MEDOIDS algorithm etc.；There are many sample characteristics of every profession and trade, provide information abundant,But the workload of data acquisition is also increased to a certain extent, it is often more important that in many situations, between many variablesThere may be correlations, to increase the complexity of case study, if analyzed respectively each index, analysis is oftenBe it is isolated, the information in data cannot be fully utilized, therefore blindly reduce index to lose many useful information, to produceThe conclusion of mistake is given birth to, dimensionality reduction mode can have: singular value decomposition (SVD), principal component analysis (PCA), factorial analysis (FA), independenceConstituent analysis (ICA) etc., using abnormal test Rating Model as a part of air control model, without being marked to sample dataNote, the client being rejected can also be used to solve positive and negative sample imbalance as sample data often without labelProblem.
Step S280 obtains air control model by enterprise's Rating Model in conjunction with abnormal test Rating Model.
Wherein, air control model can will need the enterprise characteristic for carrying out business risk assessment to input enterprise's Rating Model respectivelyIt is exported with abnormal test Rating Model, the business risk scoring and abnormal test Rating Model for obtaining the output of enterprise's Rating ModelBusiness risk scoring, the business risk of business risk scoring and the output of abnormal test Rating Model that enterprise's Rating Model is exportedScoring is averaged, and average value is the final business risk scoring of air control model output.
In above-mentioned air control model modelling approach, positive sample data set and negative sample data set, validation data set are obtained, will be testedCard data set, positive sample data set and negative sample data set input machine learning model to be trained and are trained, until meetingAfter training condition, enterprise's Rating Model is obtained, is examined using validation data set, positive sample data set and negative sample data set as abnormalThe sample characteristics of Rating Model are tested, abnormal test Rating Model is obtained；By enterprise's Rating Model and abnormal test Rating Model knotIt closes, obtains air control model, carrying out business risk scoring in conjunction with abnormal test Rating Model reduces resultant error, further improvesAir control model accuracy rate and stability.
In one embodiment, validation data set, positive sample data set and negative sample data set are inputted into machine to be trainedDevice learning model is trained, until after meeting training condition, the step of obtaining enterprise's Rating Model, comprising: by positive sample numberMachine learning model to be trained is inputted with negative sample data set according to collection；Machine learning model to be trained is based on positive sample dataCollection and negative sample data set, the machine learning model according to corresponding data characteristics vector, after being trained；By validation data setMachine learning model after input training scores, and obtains business risk scoring；When business risk scoring within a preset range,Meet training condition, obtains enterprise's Rating Model.
Wherein, machine learning model to be trained is by the positive sample data set of input and negative sample data set, by each positive sampleData in notebook data collection and negative sample data set carry out characteristic vector pickup and are learnt according to corresponding data characteristics vectorTo corresponding data characteristics vector, machine learning model after being trained, by validation data set to the engineering after trainingIt practises model to be verified, the feature vector for the data that the machine learning model after training is concentrated based on verify data determines verifyingThe feature vector of data in data set is similar, or the and negative sample that carries out feature vector with the data in positive sample data setData progress feature vector in data set is similar, the corresponding business risk scoring of each data that output verify data is concentrated, enterpriseAfter industry risk score meets training condition, enterprise's Rating Model, the corresponding business risk of each data that verify data is concentrated are obtainedWhen scoring within a preset range, meet training condition, preset range is that user indicates according to each data that each verify data is concentratedEnterprise practical risk situation determine, the machine learning model after training is verified by validation data set, furtherImprove air control model accuracy rate.
In one embodiment, it is commented using validation data set, positive sample data set and negative sample data set as abnormal testThe sample characteristics of sub-model, obtain abnormal test Rating Model the step of, comprising: by validation data set, positive sample data set andNegative sample data set carries out feature extraction as the sample characteristics of abnormal test Rating Model, obtain the features of each sample characteristics toAmount；Each feature vector is clustered, each group variety is obtained；It is distributed according to the feature space of each group variety, determines tribute in each group varietyOffer maximum sample characteristics；Using the maximum sample characteristics of contribution degree as the assessment feature of air control model, obtains abnormal test and commentSub-model.
Wherein, after establishing abnormal test Rating Model further include: using in industry all characteristic M and enterprise it is extremely specialBusiness risk scoring is modified greater than the characteristic P of average level in sign, obtains abnormal test Rating Model；Specifically:All enterprise characteristic numbers of input abnormal test Rating Model are referred to by all characteristic M in industry, all characteristic MAmount；Enterprise's off-note is analyzed, determines the characteristic P for being greater than average level in each enterprise's off-note, enterprise is abnormalFeature refers to the enterprise's off-note exported by abnormal test Rating Model；According to all characteristic M in industry and greatlyIn the characteristic P of average level, determine that the characteristic accounting of enterprise, the characteristic accounting of enterprise refer to enterprise off-note PAccounting in all characteristic M；According to the difference of the characteristic accounting of enterprise and intermediate accounting, the deviation journey of enterprise is determinedDegree, intermediate accounting refer to that integral level accounting placed in the middle in the industry, intermediate accounting are 0.5；By the departure degree of enterprise withBusiness risk scoring is added, and obtains final business risk scoring；Such as: final business risk probability are as follows:
M: all characteristics of specific enterprise in industry.P: it is greater than the characteristic of average level in enterprise's off-note.The characteristic accounting of enterprise, intermediate accounting are 0.5.The departure degree of enterprise, score: λ: business risk scoring is adjustedThe hyper parameter of the departure degree of score and enterprise is saved, user can adjust the departure degree of λ enterprise according to the actual situation.Pass through baseThe preliminary business risk probability is modified in enterprise's off-note, improves the accuracy of business risk assessment result.
Abnormal test model can will be greater than the feature letter of the characteristic P or the characteristic P less than average level of average levelBreath be collectively labeled as exception, for improve enterprise departure degree precision, when enterprise's off-note is both less than the feature of average levelWhen, the characteristic accounting of enterprise is subtracted by intermediate accounting, determines the preliminary departure degree of enterprise；When the value of preliminary departure degreeWhen greater than 0, preliminary departure degree is determined as to the departure degree of enterprise, when the value of preliminary departure degree is less than 0, enterprise it is inclinedIt is 0 from degree, such as:
Abnormal test model can will be greater than the feature letter of the characteristic P or the characteristic P less than average level of average levelBreath be collectively labeled as exception, for improve enterprise departure degree precision, when enterprise's off-note is both less than the feature of average levelWhen, the characteristic accounting of enterprise is subtracted by intermediate accounting, determines the preliminary departure degree of enterprise；When the value of preliminary departure degreeWhen greater than 0, preliminary departure degree is determined as to the departure degree of enterprise, when the value of preliminary departure degree is less than 0, enterprise it is inclinedIt is 0 from degree, such as:
In one embodiment, by enterprise's Rating Model in conjunction with abnormal test Rating Model, the step of air control model is obtainedSuddenly, comprising: the second enterprise based on the first business risk scoring of enterprise's Rating Model output with the output of abnormal test Rating ModelThe mode that industry risk score is averaged is combined, and obtains air control model.
Wherein, air control model can will need the enterprise characteristic for carrying out business risk assessment to input enterprise's Rating Model respectivelyWith abnormal test Rating Model, the business risk scoring (scoring of the first business risk) and exception of the output of enterprise's Rating Model are obtainedThe business risk of Rating Model output is examined to score (scoring of the second business risk), the business risk that enterprise's Rating Model is exportedThe business risk scoring of scoring and the output of abnormal test Rating Model is averaged, and average value is the final of air control model outputBusiness risk scoring.
In one embodiment, the step of obtaining positive sample data set and negative sample data set, validation data set includes: to obtainEnterprise's financial data sample, invoice information sample, the enterprise operation data sample of Qu Ge enterprise；By enterprise's financial data sample,Invoice information sample, enterprise operation data sample are based on data label and are matched with positive negative sample inventory, obtain positive sample numberAccording to collection and negative sample data set；The financial data sample, invoice information sample, enterprise operation data sample of non-successful match are madeFor validation data set.
Wherein, what positive sample inventory referred to belonging to positive sample data is collected and formulates corresponding label, negative sampleWhat this inventory referred to belonging to negative sample data is collected and formulates corresponding label, obtains the business finance number of each enterpriseIt needs each sample data carrying out positive negative sample division according to sample, invoice information sample, enterprise operation data sample, can be based onData label can be matched by positive negative sample inventory, positive sample data set and negative sample data set be obtained, such as positive and negativeNot having the data of data label to belong to uncertain in sample inventory is therefore positive sample or negative sample can be used as verifying numberAccording to the data of concentration, no label data can also be used in the foundation of air control model, client is solved and be rejected and loseThe problem of negative sample.
In one embodiment, as shown in figure 3, providing a kind of business risk appraisal procedure, the method includes the stepsS420 to step S440:
Step S420 obtains enterprise's financial data, the invoice information, enterprise operation data of enterprise to be scored.
Wherein, the enterprise's financial data of enterprise to be scored, invoice information, enterprise operation data can be obtained to database,It is also possible to user to input by terminal, the enterprise and each side that enterprise's financial data is embodied by Funds Movement in the process of reproductionThe data of the economic relation in face；Invoice refers to all entity and individual in purchasing and selling commodities, offer or receives service and be engaged inIn other business activities, the business voucher issued and collected, invoice information refers to all information on invoice, such as: the timeInformation, invoice codes information, affiliated enterprise's information, tax items information etc.；Enterprise operation data refer to that the enterprise summarizes and commentsThe analysis indexes of valence financial position of the enterprise and management performance, including debt paying ability index, operation ability index, Profitability IndexWith developing ability index.
Enterprise's financial data, invoice information, enterprise operation data input air control model are carried out business risk by step S440Assessment obtains business risk assessment result.
Wherein, enterprise's financial data, invoice information, enterprise operation data are inputted into air control model, the enterprise of air control modelRating Model is analyzed according to enterprise's financial data, invoice information, enterprise operation data, obtains the scoring of the first business risk；Abnormal test Rating Model is analyzed according to enterprise's financial data, invoice information, enterprise operation data, and output business risk is commentedPoint and enterprise's off-note, obtain all characteristics in the affiliated industry of the enterprise；Enterprise's off-note is analyzed, is determinedIt is greater than the characteristic of average level in each enterprise's off-note；According to all characteristics in industry and greater than the spy of average levelNumber is levied, determines the characteristic accounting of enterprise；According to the difference of the characteristic accounting of enterprise and intermediate accounting, the deviation of enterprise is determinedDegree；The departure degree of enterprise is added with preliminary business risk probability, obtains the scoring of the second business risk；It obtains the first enterpriseThe average value of industry risk score and the scoring of the second business risk, which is exported as business risk assessment result.
Above-mentioned business risk appraisal procedure, by the enterprise's financial data, invoice information, the enterprise's warp that obtain enterprise to be scoredSeek data；The air control model of enterprise's financial data, invoice information, the input of enterprise operation data is subjected to business risk assessment, is obtainedBusiness risk assessment result, air control model score enterprise's financial data, invoice information, enterprise operation data mould by enterpriseType carries out business risk scoring in conjunction with abnormal test Rating Model reduces resultant error, improves the accurate of business risk assessmentRate.
It should be understood that although each step in the flow chart of Fig. 2-3 is successively shown according to the instruction of arrow,These steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly stating otherwise herein, these stepsExecution there is no stringent sequences to limit, these steps can execute in other order.Moreover, at least one in Fig. 2-3Part steps may include that perhaps these sub-steps of multiple stages or stage are not necessarily in synchronization to multiple sub-stepsCompletion is executed, but can be executed at different times, the execution sequence in these sub-steps or stage is also not necessarily successivelyIt carries out, but can be at least part of the sub-step or stage of other steps or other steps in turn or alternatelyIt executes.
In one embodiment, as shown in figure 4, providing a kind of air control model modeling device, comprising: data set obtains mouldBlock 310, enterprise's Rating Model training module 320, abnormal test Rating Model establish module 330 and models coupling module 340,In:
Data set acquisition module 310, for obtaining positive sample data set and negative sample data set, validation data set；
Enterprise's Rating Model training module 320, for validation data set, positive sample data set and negative sample data set is defeatedEnter machine learning model to be trained to be trained, until obtaining enterprise's Rating Model after meeting training condition；
Abnormal test Rating Model establishes module 330, is used for validation data set, positive sample data set and negative sample dataCollect the sample characteristics as abnormal test Rating Model, obtains abnormal test Rating Model；
Models coupling module 340, for enterprise's Rating Model in conjunction with abnormal test Rating Model, to be obtained air control mouldType.
In one embodiment, enterprise's Rating Model training module 320 includes: data input cell, is used for positive sampleData set and negative sample data set input machine learning model to be trained；Enterprise's Rating Model training unit, for wait trainMachine learning model be based on positive sample data set and the negative sample data set, according to corresponding data characteristics vector, obtainMachine learning model after training；Authentication unit, for commenting the machine learning model after validation data set input trainingPoint, obtain business risk scoring；Judging unit meets training condition, obtains for scoring within a preset range when business riskEnterprise's Rating Model.
In one embodiment, it includes: characteristic vector pickup unit that abnormal test Rating Model, which establishes module 330, is used forFeature is carried out using validation data set, positive sample data set and negative sample data set as the sample characteristics of abnormal test Rating ModelIt extracts, obtains the feature vector of each sample characteristics；Cluster cell obtains each group variety for clustering each feature vector；SampleEigen determination unit is determined to contribute maximum sample characteristics in each group variety for being distributed according to the feature space of each group variety；Characteristics determining unit is assessed, for obtaining abnormal inspection using the maximum sample characteristics of contribution degree as the assessment feature of air control modelTest Rating Model.
In one embodiment, models coupling module 340 includes being used for: the first enterprise based on the output of enterprise's Rating ModelThe mode that the second business risk scoring that risk score is exported with abnormal test Rating Model is averaged is combined, and obtains windControl model.
In one embodiment, data set acquisition module 310 includes being used for: obtaining the enterprise's financial data sample of each enterpriseSheet, invoice information sample, enterprise operation data sample；By enterprise's financial data sample, invoice information sample, enterprise operation dataSample is based on data label and is matched with positive negative sample inventory, obtains positive sample data set and negative sample data set；It will notWith successful financial data sample, invoice information sample, enterprise operation data sample as validation data set.
In one embodiment, as shown in figure 5, providing a kind of business risk assessment device, comprising:
Data acquisition module 510, for obtaining enterprise's financial data, invoice information, the enterprise operation number of enterprise to be scoredAccording to；
Business risk grading module 520, the wind for inputting enterprise's financial data, invoice information, enterprise operation dataIt controls model and carries out business risk scoring, obtain business risk value.
Specific about air control model modeling device limits the limit that may refer to above for air control model modelling approachFixed, specific about business risk assessment device limits the restriction that may refer to above for business risk appraisal procedure,This is repeated no more.Modules in above-mentioned air control model modeling device, business risk assessment device can be fully or partially throughSoftware, hardware and combinations thereof are realized.Above-mentioned each module can be embedded in the form of hardware or independently of the place in computer equipmentIt manages in device, can also be stored in a software form in the memory in computer equipment, in order to which processor calls execution or moreThe corresponding operation of modules.
In one embodiment, a kind of computer equipment is provided, which can be server, internal junctionComposition can be as shown in Figure 6.The computer equipment include by system bus connect processor, memory, network interface andDatabase.Wherein, the processor of the computer equipment is for providing calculating and control ability.The memory packet of the computer equipmentInclude non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program and dataLibrary.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculatingThe database of machine equipment is for storing the data such as sample characteristics.The network interface of the computer equipment is used for logical with external terminalCross network connection communication.To realize a kind of air control model modeling, business risk assessment when the computer program is executed by processorMethod.
It will be understood by those skilled in the art that structure shown in Fig. 6, only part relevant to application scheme is tiedThe block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme, specific computer equipmentIt may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.
In one embodiment, a kind of computer equipment, including memory and processor are provided, which is stored withComputer program, the processor perform the steps of when executing computer program
Obtain positive sample data set, negative sample data set and validation data set；By validation data set, positive sample data set andNegative sample data set inputs machine learning model to be trained and is trained, until obtaining enterprise's scoring after meeting training conditionModel；Using validation data set, positive sample data set and negative sample data set as the sample characteristics of abnormal test Rating Model, obtainObtain abnormal test Rating Model；By enterprise's Rating Model in conjunction with abnormal test Rating Model, air control model is obtained.
In one embodiment, processor execute computer program when also perform the steps of by positive sample data set withNegative sample data set inputs machine learning model to be trained；Machine learning model to be trained is based on positive sample data set and bearsSample data set, the machine learning model according to corresponding data characteristics vector, after being trained；Validation data set is inputted and is instructedMachine learning model after white silk scores, and obtains business risk scoring；When business risk scores within a preset range, satisfaction is instructedThe condition of white silk obtains enterprise's Rating Model.
In one embodiment, it also performs the steps of when processor executes computer program by validation data set, positive sampleThe sample characteristics progress feature extraction of notebook data collection and negative sample data set as abnormal test Rating Model, it is special to obtain each sampleThe feature vector of sign；Each feature vector is clustered, each group variety is obtained；It is distributed, is determined according to the feature space of each group varietyMaximum sample characteristics are contributed in each group variety；Using the maximum sample characteristics of contribution degree as the assessment feature of air control model, obtainAbnormal test Rating Model.
In one embodiment, it also performs the steps of when processor executes computer program based on enterprise's Rating ModelThe mode that the first business risk scoring of output is averaged with the second business risk scoring of abnormal test Rating Model outputIt is combined, obtains air control model.
In one embodiment, the enterprise for obtaining each enterprise is also performed the steps of when processor executes computer programFinancial data sample, invoice information sample, enterprise operation data sample；By enterprise's financial data sample, invoice information sample, enterpriseIndustry is managed data sample and is matched based on data label with positive negative sample inventory, and positive sample data set and negative sample data are obtainedCollection；Using the financial data sample of non-successful match, invoice information sample, enterprise operation data sample as validation data set.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, is calculatedMachine program performs the steps of when being executed by processor
Obtain positive sample data set, negative sample data set and validation data set；By validation data set, positive sample data set andNegative sample data set inputs machine learning model to be trained and is trained, until obtaining enterprise's scoring after meeting training conditionModel；Using validation data set, positive sample data set and negative sample data set as the sample characteristics of abnormal test Rating Model, obtainObtain abnormal test Rating Model；By enterprise's Rating Model in conjunction with abnormal test Rating Model, air control model is obtained.
In one embodiment, it is also performed the steps of when computer program is executed by processor by positive sample data setMachine learning model to be trained is inputted with negative sample data set；Machine learning model to be trained be based on positive sample data set withNegative sample data set, the machine learning model according to corresponding data characteristics vector, after being trained；Validation data set is inputtedMachine learning model after training scores, and obtains business risk scoring；When business risk scoring within a preset range, meetTraining condition obtains enterprise's Rating Model.
In one embodiment, it is also performed the steps of by validation data set, just when computer program is executed by processorThe sample characteristics progress feature extraction of sample data set and negative sample data set as abnormal test Rating Model, obtains each sampleThe feature vector of feature；Each feature vector is clustered, each group variety is obtained；It is distributed, is determined according to the feature space of each group varietyMaximum sample characteristics are contributed in each group variety out；Using the maximum sample characteristics of contribution degree as the assessment feature of air control model, obtainObtain abnormal test Rating Model.
In one embodiment, it is also performed the steps of when computer program is executed by processor based on enterprise's scoring mouldThe side that the first business risk scoring of type output is averaged with the second business risk scoring of abnormal test Rating Model outputFormula is combined, and obtains air control model.
In one embodiment, the enterprise for obtaining each enterprise is also performed the steps of when computer program is executed by processorIndustry financial data sample, invoice information sample, enterprise operation data sample；By enterprise's financial data sample, invoice information sample,Enterprise operation data sample is based on data label and is matched with positive negative sample inventory, obtains positive sample data set and negative sample numberAccording to collection；Using the financial data sample of non-successful match, invoice information sample, enterprise operation data sample as validation data set.
In one embodiment, a kind of computer equipment, including memory and processor are provided, which is stored withComputer program, the processor execute computer program when perform the steps of obtain enterprise to be scored enterprise's financial data,Invoice information, enterprise operation data；By enterprise's financial data, invoice information, the input of enterprise operation data the air control modelBusiness risk assessment is carried out, business risk assessment result is obtained.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, is calculatedMachine program performs the steps of the enterprise's financial data for obtaining enterprise to be scored, invoice information, enterprise's warp when being executed by processorSeek data；Enterprise's financial data, invoice information, the input of enterprise operation data the air control model are carried out business risk to commentEstimate, obtains business risk assessment result.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be withRelevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computerIn read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,To any reference of memory, storage, database or other media used in each embodiment provided herein,Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may includeRandom access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancingType SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of above embodiments can be combined arbitrarily, for simplicity of description, not to above-described embodimentIn each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lanceShield all should be considered as described in this specification.
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneouslyIt cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the artIt says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the applicationRange.Therefore, the scope of protection shall be subject to the appended claims for the application patent.