CN109961296A - Merchant type recognition methods and device - Google Patents
Merchant type recognition methods and device Download PDFInfo
- Publication number
- CN109961296A CN109961296A CN201711416934.3A CN201711416934A CN109961296A CN 109961296 A CN109961296 A CN 109961296A CN 201711416934 A CN201711416934 A CN 201711416934A CN 109961296 A CN109961296 A CN 109961296A
- Authority
- CN
- China
- Prior art keywords
- trade company
- information
- identified
- merchant
- type
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06Q30/00—Commerce
- G06Q30/018—Certifying business or products
- G06Q30/0185—Product, service or business identity fraud
Landscapes
- Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Finance (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of merchant type recognition methods and devices, belong to data mining analysis field.The merchant type recognition methods includes: to receive merchant type identification request;Obtain information and the userspersonal information of trade company to be identified corresponding with merchant type identification request;The information and the corresponding characteristic of the userspersonal information for extracting the trade company to be identified respectively, obtain the predictive data set of the trade company to be identified;The predictive data set is input to trained Merchant Category prediction model, obtains the output result of the Merchant Category prediction model;And exporting as a result, determining the type of the trade company to be identified according to the Merchant Category prediction model.Technical solution of the present invention combines data mining analysis technology, it is the type that can recognize trade company by simple pretreatment operation and Merchant Category prediction model, reduce nonterminal character obtain needed for human cost, model deployment building complexity it is low, model it is versatile.
Description
Technical field
The present invention relates to data mining analysis field, in particular to a kind of merchant type recognition methods and device.
Background technique
It when currently, carrying out network trading between trade company and individual consumer is realized by third-party platform mostly.Third
The service provider of Fang Pingtai can provide advanced interface, complete payment application, technological development, equipment debugging, activity marketing etc. for trade company
Full ecological chain service.Service provider, also can be to specific type trade company when carrying out the various marketing activities for particular merchant type
Carry out commission preferential rate.For example, 0 rate for Wei Can trade company of wechat payment development is preferential and respective service quotient returns servant
Activity.In this activity, wechat payment needs to subsidize Wei Can trade company accordingly.In the case where budget is fixed,
If the Wei Can trade company of personation can not be identified, these personation trade companies will abduct a large amount of subsidies of wechat, so that budget shifts to an earlier date quilt
It is finished.Furthermore the personnel that this 0 rate activity can may also be originally engaged in the illegal activities such as brush list, money laundering utilize.And
And in the businessman by wechat payment gathering, there is also developments to violate state's laws, the business for needing to be hit, for example washes
Money, swindle is pornographic, gambling etc..Therefore, precisely identify merchant type for the platform income of guarantee service provider and marketing activity
The normal use of budget is of crucial importance, precisely identifies that merchant type is an extremely important problem, and recognition result has
Extensive practical application, such as big data analysis, advertisement dispensing etc..
Existing technical solution is classified using manual identified, or uses conventional machines learning classification model, is needed a large amount of
Artificial participation, and due to various reasons, the type of most trade companies carries out trade company's payment account note in third-party platform
When volume, there is no manual confirmation and verification is passed through, cause service provider that can not directly learn the actual types of trade company.
Summary of the invention
In order to solve problems in the prior art, the present invention provides a kind of merchant type recognition methods and devices, according to letter
Single pretreatment operation and Merchant Category prediction model can precisely identify the type of trade company.The technical solution is as follows:
On the one hand, the present invention provides a kind of merchant type recognition methods, which comprises
Receive merchant type identification request;
Obtain information and the userspersonal information of trade company to be identified corresponding with merchant type identification request;
The information and the corresponding characteristic of the userspersonal information for extracting the trade company to be identified respectively, obtain institute
State the predictive data set of trade company to be identified;
The predictive data set is input to trained Merchant Category prediction model, it is pre- to obtain the Merchant Category
Survey the output result of model;And
According to the output of the Merchant Category prediction model as a result, determining the type of the trade company to be identified.
On the other hand, the present invention provides a kind of merchant type identification device, described device includes:
Request receiving module, for receiving merchant type identification request;
Data obtaining module, for obtaining the information of trade company to be identified corresponding with merchant type identification request
And userspersonal information;
Characteristic extracting module, the information and the userspersonal information for extracting the trade company to be identified respectively are corresponding
Characteristic, obtain the predictive data set of the trade company to be identified;
Data input module is obtained for the predictive data set to be input to trained Merchant Category prediction model
To the output result of the Merchant Category prediction model;
Trade company's identification module, for the output according to the Merchant Category prediction model as a result, determining the quotient to be identified
The type at family.
Technical solution bring provided by the invention has the beneficial effect that:
1) versatile, it may be convenient to applied in different business scenarios;
2) model performance is high, and maintenance cost is low;
3) it significantly improves and doubtful fraud trade company is recalled, and can be balanced and be recalled according to business side's manpower-accurately referred to
Mark;
4) users personal data abundant is made full use of.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, required in being described below to embodiment
The attached drawing used is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention,
For those of ordinary skill in the art, without creative efforts, it can also be obtained according to these attached drawings
Other attached drawings.
Fig. 1 is the flow chart of merchant type recognition methods provided in an embodiment of the present invention;
Fig. 2 is the training method flow chart of merchant type model provided in an embodiment of the present invention;
Fig. 3 is the structural map of the stacking model in trade company's prediction disaggregated model provided in an embodiment of the present invention;
Fig. 4 is the acquisition of information side of sample trade company in the training method of merchant type model provided in an embodiment of the present invention
The flow chart of method;
Fig. 5 is the module frame chart of merchant type identification device provided in an embodiment of the present invention;
Fig. 6 is the flow diagram of trade company's prediction disaggregated model training and practice provided in an embodiment of the present invention;
Fig. 7 is the hardware block diagram of the terminal of merchant type identification device provided in an embodiment of the present invention.
Specific embodiment
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
It is the embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, the common skill in this field
Art personnel every other embodiment obtained without making creative work, all should belong to protection of the present invention
Range.
It should be noted that description and claims of this specification and term " first " in above-mentioned attached drawing, "
Two " etc. be to be used to distinguish similar objects, without being used to describe a particular order or precedence order.It should be understood that making in this way
Data are interchangeable under appropriate circumstances, so that the embodiment of the present invention described herein can be in addition to scheming herein
Sequence other than those of showing or describe is implemented.In addition, term " includes " and " having " and their any deformation, it is intended that
Be to cover it is non-exclusive include, for example, containing the process, method of a series of steps or units, device, product or equipment
Those of be not necessarily limited to be clearly listed step or unit, but may include be not clearly listed or for these processes,
The intrinsic other step or units of method, product or equipment.In one embodiment of the invention, a kind of trade company's class is provided
Type recognition methods, referring to Fig. 1, method flow includes:
S101, merchant type identification request is received.
Specifically, identification request is initiated, the request for certain a kind of trade company (such as fraud trade company, Wei Can trade company etc.)
Purpose be identification obtain the trade company type (merchant type can on line/line under, enclose meal/non-enclose meal, cheat/normally with
And other types etc.), it initiates request and needs trigger action, in this application, trigger action may be by user in related application
Triggering when checking the merchant information in (such as social application, map, shopping platform), can also be by the front end management of service provider
Personnel carry out triggering by hand and either carry out regular or irregular automatic trigger, the present invention by the background system of service provider
The mode for initiating request to triggering is not especially limited.
S102, the information and individual subscriber for obtaining trade company to be identified corresponding with merchant type identification request
Information.
Specifically, the ID (full name are as follows: Virtual of trade company to be identified can be attached in merchant type identification solicited message
Identity electronic identification, i.e. universal account, are commonly called as network identification card) information, trade company ID letter
Breath is can to find corresponding trade company according to id information to the unique encodings of merchant identification mark.Merchant type identification request is opposite
The information for the trade company to be identified answered includes transaction journal information and trade company's id information of trade company to be identified etc., wherein transaction flow
Water number evidence, including but not limited to: a) exchange hour;B) the transaction address ip;C) transaction amount section;D) channel of disbursement (debit
Card, credit card, change);E) pay scene (app is jumped, barcode scanning payment, jsapi etc.);F) information such as successfully whether are paid.
Specifically, the personal information of user refers to user at related application (such as social application, map, shopping platform)
In personal information, userspersonal information can be the personal information of wechat user in the present embodiment, including but not limited to: a)
Age;B) gender;C) permanent provinces and cities;D) equipment (ios, andriod) is used;E) it registers source (cell-phone number, No. qq) etc., is
Protection privacy, above section data can reduce accuracy, for example exact amount is transformed into less than 50,50-100,
The section 100-500,500-1000, etc..
In the present embodiment, server or terminal can obtain the merchant type respectively from corresponding database
Information and the userspersonal information of corresponding trade company to be identified are requested in identification, such as obtain from wechat trade company collective database
The information for taking trade company to be identified obtains the personal information etc. of user from wechat users personal data library.
Specifically, userspersonal information may include in related application (such as social application, map, shopping platform)
The information that user and trade company to be identified trade also may include the trade company to be identified in the information of trade company to be identified and use
The information that family is traded in related application (such as social application, map, shopping platform), user trade with trade company
Information can directly in the transaction journal acquisition of information of trade company to be identified, can also by searching for the trade company to be identified into
The ID of the user of row transaction, obtains to obtain by User ID from corresponding userspersonal information.By obtaining and trade company's class
Information and the userspersonal information of corresponding trade company to be identified are requested in type identification, form the prediction of merchant type to be identified
Original data source.
S103, the information for extracting the trade company to be identified respectively and the corresponding characteristic of the userspersonal information, obtain
To the predictive data set of the trade company to be identified.
Specifically, the information of trade company to be identified includes the transaction journal information of the trade company to be identified, by from S102
The feature field extracted in the original data source of the prediction of merchant type to be identified is as the information of trade company to be identified and user
The corresponding characteristic of people's information, these characteristics include but is not limited to:
1. transaction size: amount of money distribution, stroke count distribution, number of days, account number, payment failure rate etc.;
2. user characteristics: age, gender, repeat buying register source, using equipment, permanent provinces and cities etc.;
3. exchange hour: transaction accounting in different time periods;
4. paying denomination: the transaction accounting in different amount of money sections;Accounting highest transaction section;
5. the means of payment: channel, scene etc..
For the characteristic that these are made of feature field, need to be converted into the feature vector of computer capacity reading, because
It can not directly be used by model for nonumeric feature, need that they are become numerical characteristics by way of transcoding.Such as: it uses
00001 indicates the maximum trade company of this transaction size, and 00010 indicates second largest trade company of this transaction size, to user spy
The transcoding mode of sign etc. is also similar.It can be the feature field of identification merchant type, such as name, property by extracting some
Not, repeat buying, registration source etc., according to these feature fields, from the information and individual subscriber of above-mentioned trade company to be identified
The characteristic of corresponding feature field composition as above is obtained in information respectively, and characteristic is changed into feature vector, this
A little feature vectors form the predictive data set of the trade company to be identified.
S104, the predictive data set is input to trained Merchant Category prediction model, obtains the trade company point
The output result of class prediction model.
Specifically, trained Merchant Category prediction model is obtained according to training dataset training, for different
Business scenario, training dataset is different, and the merchant type identified is not also identical, such as in the business scenario of fraud identification,
Training dataset is using characteristic relevant to fraud identification, then the Merchant Category prediction model can recognize that quotient
Family type be to cheat trade company or being normal trade company, such as in the business scenario of solid shop/brick and mortar store Classification and Identification, training data centralized procurement
It is characteristic relevant to solid shop/brick and mortar store classification, then the Merchant Category prediction model can recognize that merchant type can be with
For hospital, supermarket, gas station etc., i.e., mould is predicted by the way that predictive data set is input to trained different Merchant Category
Type obtains the corresponding output result of different Merchant Category prediction models.
S105, exporting as a result, determining the type of the trade company to be identified according to the Merchant Category prediction model.
Specifically, Merchant Category prediction model is more classification prediction models, and the quantity of the classification results of output can be two
It plants and two or more.Before the training Merchant Category prediction model, setting model output result first is corresponding with merchant type
Rule, the classification results for exporting model according to rule are corresponding to mark related merchant type, to obtain the class of trade company to be identified
Type.
The present invention is by the extraction to trade company's flowing water information and userspersonal information's progress characteristic information, to extraction
Feature carries out simple pretreatment operation, and characteristic is input to Merchant Category prediction model by treated, does not need artificial
Confirmation and verification, can accurately identify the type of trade company, reduce the human cost needed for nonterminal character obtains, the portion of model
Administration building complexity it is low, model it is versatile.
In one embodiment of the invention, a kind of training method of Merchant Category prediction model is provided, referring to fig. 2,
Method flow includes:
S21, the information for obtaining sample trade company and userspersonal information.
Specifically, the information of sample trade company includes the id information of sample trade company, the merchant type mark for being labeled with sample trade company
The information of label and the transaction journal of the sample trade company.In userspersonal information and above-mentioned identification process in training process
The acquisition source of userspersonal information is identical, is all that in related application, (such as social application, map, shopping are flat by obtaining user
Platform etc.) in personal information.Userspersonal information can be the personal information of wechat user in the present embodiment, including but unlimited
In: a) age;B) gender;C) permanent provinces and cities;D) equipment (ios, andriod) is used;E) source (cell-phone number, No. qq) is registered
Deng in order to protect privacy, above section data can reduce accuracy, for example exact amount is transformed into less than 50,50-
The section 100,100-500,500-1000, etc..
Specifically, the specific acquisition process of the information of sample trade company includes:
The information of S210, the training samples information for obtaining target trade company respectively and all trade companies in related application.
Specifically, the training samples information of target trade company includes the id information and corresponding remarks label of the target trade company,
Remarks label is the corresponding merchant type of target trade company manually marked, and remarks label is the concrete type of trade company, remarks
The mode of label has very much, and specific merchant type is defined according to specific business scenario.Such as in fraud identification,
Normal trade company and fraud trade company can be divided into.The logic of identification merchant type is purely determined by business scenario.If you do not need to fraud
Identification, would not carry out fraud identification.In the classification for certain solid shop/brick and mortar store, hospital, supermarket, gas station etc. can be divided into.Institute
The Transaction Information for stating all trade companies in related application includes the id information and transaction journal information of all trade companies, on
It states information and is present in related application (such as social application, map, shopping platform).It, first will be to instruction before being trained
Experienced data are pre-processed, and determine sample trade company and its relevant information.It needs from the merchant type by manually marking
Selected part trade company is as target trade company, the quotient to the target trade company that by the id information of the target trade company and manually marks
Training samples information of the remarks label of family type as the target trade company, in addition, it is also necessary to pull related application (such as social activity
Using, map, shopping platform etc.) in all trade company's Transaction Informations, trade company's Transaction Information includes all trade company ID letters
Breath and transaction journal information.
S211, the Transaction Information for traversing all trade companies, by the id information of all trade companies and the target trade company
Id information matched, the trade company of successful match in all trade companies is determined as sample trade company, and by the remarks mark
Sign the merchant type label as the sample trade company.
Specifically, target trade company id information is matched with all trade company's id informations in related application, matching at
Trade company in the related application is determined as sample trade company by function, and stamp merchant type of the remarks label as sample trade company
Label, using the corresponding trade company's Transaction Information of the sample trade company and merchant type label as the information of the sample trade company, i.e.,
The information of sample trade company includes the corresponding trade company's Transaction Information of the sample trade company and the corresponding trade company, target trade company that matching obtains
Type label.
In one embodiment of the present of invention, since the logic of identification merchant type is purely determined by business scenario, if needed
Identification is cheated, just will do it fraud identification.If you do not need to fraud identification, would not carry out fraud identification.Merchant type mark
Label can include but is not limited to following a few classes: on line/line under, enclose meal/non-and enclose meal, fraud/normally and other types label
Deng the relevant situation for the identification of this few class merchant type includes:
1. on line/identification of Xian Xia trade company.In certain business scenarios, need to determine whether a trade company has a line
Lower solid shop/brick and mortar store, since there are some apparent differences, such as quotient on pure line in the trade company on pure line and the trade company with solid shop/brick and mortar store under line
The place of family transaction and Annual distribution all relative distributions, and the place of the transaction of the trade company of solid shop/brick and mortar store is fixed, transaction under line
Time is also usually the normal business hours for belonging to trade company, by obtaining the place traded and temporal information and according to correspondence
The characteristics of carry out on line/the identification of Xian Xia trade company.
2. enclosing meal/Fei Weican trade company identification.Wei Can trade company, which refers to, provides more people businessman for having a dinner party service in catering trade, lead to
It is often first just to check after the meal, does not include fast food, snack is baked and banked up with earth.In the preferential activity for " enclosing meal " trade company, essence is needed
Really identification, which is pretended to be, encloses the trade company that meal enjoys privileges.Wei Can trade company and other food and drink trade companies or other any type trade companies have one
A little differences, for example have dinner and be generally focused on the specific period, the distribution etc. of transaction amount.
3. multi-class trade company's identification can be divided into hospital, supermarket, gas station etc. in the classification for certain solid shop/brick and mortar store
Merchant type.
The present invention is by combining merchant information and userspersonal information to establish accordingly for different business application scenarios
The accurate identification model of merchant type is the type that can recognize trade company by simple pretreatment operation and Merchant Category prediction model,
Reduce nonterminal character obtain needed for human cost, model deployment building complexity it is low, model it is versatile.
S213, userspersonal information described in the related application is obtained.
Specifically, server or terminal can obtain correlation described in S210 respectively from corresponding database and answer
The information of trade company in and userspersonal information.By taking related application is wechat as an example, server or terminal are deposited from wechat
The information for obtaining all trade companies in the database of merchant information is put, the individual of user is obtained from wechat users personal data library
Information etc..Wechat userspersonal information includes the Transaction Information of the sample trade company in wechat user and above-mentioned S210.In training
During model, server or terminal can be obtained respectively from corresponding database the transaction journal information of trade company with
And userspersonal information, the transaction journal information of trade company is obtained such as from wechat trade company collective database, from wechat individual subscriber
The personal information etc. of user is obtained in database.In this way by using the transaction journal data of wechat trade company and wechat user
People's representation data establishes the accurate identification model of corresponding merchant type for different business application scenarios, and wechat is not used only
Payment transaction pipelined data also uses wechat users personal data abundant, extracts feature, bonding machine in conjunction with business experience
Device learning classification model identifies particular merchant type to classify.
S22, the information for extracting the sample trade company respectively and the corresponding sample characteristics data of userspersonal information, obtain
To the training dataset of the Merchant Category prediction model.
Specifically, sample merchant information includes the corresponding trade company's Transaction Information of the sample trade company and merchant type label,
Userspersonal information in S22 is identical as the source of the userspersonal information in S201.Sample merchant information and individual subscriber letter
Ceasing corresponding sample characteristics data is the sample characteristics field extracted from above-mentioned sample merchant information and userspersonal information,
Including but not limited to:
1. transaction size: amount of money distribution, stroke count distribution, number of days, account number, payment failure rate etc.;
2. user characteristics: age, gender, repeat buying register source, using equipment, permanent provinces and cities etc.;
3. exchange hour: transaction accounting in different time periods;
4. paying denomination: the transaction accounting in different amount of money sections;Accounting highest transaction section;
5. the means of payment: channel, scene etc.;
6. merchant type label of sample trade company etc..
For the characteristic that these are made of feature field, need to be converted into the feature vector of computer capacity reading, because
It can not directly be used by model for nonumeric feature, need that they are become numerical characteristics by way of transcoding.Such as:
Type of service is in fraud identification scene, and the label of merchant type is normal trade company and fraud trade company, then is indicated with 00001
Normal trade company, 00010 indicates fraud trade company, and the transcoding mode to user characteristics etc. is also similar.It is some by extraction to be
Identify the feature field of merchant type, such as name, gender, repeat buying, registration source etc., according to these feature fields,
It obtains corresponding feature field as above respectively from the transaction journal information of sample trade company and userspersonal information, passes through acquisition
The information of sample trade company corresponding with merchant type identification request and userspersonal information, it is pre- to form the Merchant Category
Survey the training dataset of model.
S23, each submodel that the training dataset is respectively trained to the Merchant Category prediction model, determine institute
The parameter for stating each submodel of Merchant Category prediction model obtains the Merchant Category prediction model.
It specifically, include the submodel of multiple independent predictions in Merchant Category prediction model, it is same in order to meet the above demand
When take into account estimated performance, the technical program is using stacking model.Other machine learning classification models may also meet demand,
Here a kind of realization is only elaborated.
Specifically, stack the thought that model uses integrated model, by gather multiple independent prediction models as a result,
Reach better accuracy rate and generalization ability.The stacking model of this design scheme is made of 7 sub- models, is respectively as follows: 1. and is patrolled
It collects and returns;2. neural network;3.Gradient boosting;4. random forest; 5.Adaboosting;6.Bagging;
7.ExtraTrees.The construction for stacking model is as shown in Figure 3, wherein " x " in Fig. 3 refers to feature, described to be characterized in machine
The input of learning model.Above-mentioned seven submodels are the d1 in Fig. 3, and d2 etc., f () are the prediction results this seven submodels
The function summarized, y are final prediction result (such as the probability for belonging to Wei Can trade company).F () is used in the technical program
Aggregation function is that the probability that exports each model is added, final the result is that corresponding classification of maximum probability.
Specifically, when being trained to model, feature " x " be it is determining, by continually entering feature " x ", so as to
To be adjusted to parameter in each submodel, trade company's class of sample trade company corresponding to " y " and feature " x " so as to output
Type is close, when " y " of output is consistent with the merchant type of sample trade company corresponding to feature " x ", determines current each submodule
Parameter is the optimum state for each submodel that training obtains in type, to show that each submodel of the optimum state is quotient
The submodel of family classification prediction model.
It so far, can be by distinguishing current quotient according to business scenario using training sample data and users personal data
The type at family.In the label for the merchant type mentioned in such as above-mentioned S211, mark is the trade company of the label of " enclosing meal ", then the mould
The training sample data of type are to determine that this current merchant type is to enclose the data of meal or Fei Weican trade company, trained purpose
Precisely identifying and classify for merchant type is carried out for Merchant Category prediction model.
In one embodiment of the invention, the predictive data set is input to described in step S104 and has been trained
Merchant Category prediction model, obtain the output result of the Merchant Category prediction model, comprising:
S1041, the predictive data set is separately input to each submodel in the Merchant Category prediction model,
Obtain sub- probability corresponding to the corresponding trade company to be identified of each submodel.
Specifically, the sum of the sub- probability of each merchant type of each submodel is 1.D1 in Fig. 3, d2 to d7 etc. this
Each of seven submodels are all stand-alone trainings, and independently provide prediction result.Prediction result output be belong to it is each
The probability of classification.Such as fraud identification scene in, can be divided into normal trade company (Normal business, abbreviation NB) and
It cheats trade company (Fraudulentbusiness, abbreviation FB), feature is trained by d1 to obtain the corresponding probability of NB and FB
Respectively 0.3 and 0.7, being trained to obtain the corresponding probability of NB and FB to feature by d2 is respectively 0.4 and 0.6, i.e. d1 and
D2 is stand-alone training, and independently provides result.That is, each submodel respectively obtain be trade company to be identified trade company's class
The corresponding one group of probability of type, if the corresponding type of trade company to be identified only has 2 kinds (such as fraud/normal trade companies), every height
One group of the probability that model obtains be 2, the sum of this 2 probability be 1, if the corresponding type of trade company to be identified be N (N >
2) seed type (such as hospital, supermarket, gas station), one group of the probability that each submodel obtains are N number of, the sum of this N number of probability
It is 1.
S1042, sub- probability corresponding to the trade company to be identified is calculated by aggregation function, obtain it is described to
Identify total probability corresponding to trade company.
Specifically, aggregation function is the f () in Fig. 3, aggregation function f () be with the output result of each submodel into
The function of row " ballot " exports y.F () is divided into hard ballot and soft ballot.Hard ballot refers to that each submodel throws a ticket.Than
Such as, the target of model training is identification fraud trade company.3 are identified as cheating in 7 submodels, and fraud just has 3 tickets, 4 identifications
For non-fraud, non-fraud just has 4 tickets, so f () output is non-fraud.It is soft ballot be exactly model submodel export be each
The probability of merchant type.F () counts each classification and corresponds to the sum of probability.Probability and that maximum classification are exactly output knot
Fruit.Scikit-learnpython software library is used in modeling process, the technical detail of specific each submodel and training belong to
In the method that machine learning field is general, no longer narration in detail herein.
S1043, the total probability according to corresponding to the trade company to be identified, determine the type of the trade company to be identified.
Specifically, the embodiment of the present invention is by the way that the predictive data set is separately input to each of Merchant Category prediction
A submodel (such as d1, d2 etc.) obtains the probability of the corresponding merchant type of each submodel of the Merchant Category prediction.It is logical
Aggregation function (such as f ()) is crossed to count the probability for the corresponding merchant type of each submodel that the Merchant Category is predicted
It calculates, obtains total probability corresponding to the merchant type (such as output y), according to total probability corresponding to the merchant type, really
The type of the fixed trade company.
As an embodiment of the present invention, the type for determining trade company is by using the institute, trade company to be identified is right
The total probability answered is compared with preset threshold value, determines trade company's class of the trade company to be identified according to the result of the comparison
Type.When the sample characteristics are the characteristic information for cheating trade company, the value of the preset threshold value is between 0.6 to 1
It is described to be identified if total probability corresponding to the trade company to be identified is greater than the preset threshold value when any one value
The type of trade company is fraud trade company, if total probability corresponding to the trade company to be identified is less than the preset threshold value, institute
The type for stating trade company to be identified is normal trade company.
Specifically, such as trained model is to be applied to identify fraud trade company, when the friendship for obtaining all trade companies
After easy pipelined data and the personal data of user, corresponding feature vector predicted composition data set (feature vector composition is extracted
Predictive data set be X in Fig. 3), and the predictive data set is separately input in 7 submodels (as d1, d2 etc.)
Predicted to obtain 7 groups of corresponding probability, as d1 obtain one group of probability be (0.3,0.7), d2 obtain one group of probability be (0.4,
0.6) etc., then the 7 groups of probability obtained are calculated by aggregation function f (), as the model be to fraud trade company into
Row identification, the threshold value set are considered fraud trade company as 0.6, the i.e. trade company greater than 0.6, and on the contrary is normal trade company, if using
Hard ballot is calculated, then the recognition result of d1 is fraud trade company, and the recognition result of d2 is normal trade company, and last result is just
Being determined as merchant type by that classification of fraud and the maximum probability of normal trade company, (such as fraud, there are 3 tickets, normal trade company in trade company
There are 4 tickets, then merchant type is normal trade company).The probability of normal trade company is 0.3+0.4=if using the method for soft ballot
0.7, the probability for cheating trade company is 0.7+0.6=1.3, and since 1.3 are greater than 0.7, then it is fraud that last result, which is merchant type,
Trade company.
In the present invention, since using linear model, which provides good interpretation.Model
Output while the original value for providing the most important feature of this model and the contribution in linear model, make it easy to examine
Certain trade company of breaking is marked as some type of reason.Most important feature list is arranged by the feature importance for integrating each submodel
Sequence obtains (ignoring here can not be to the model that feature importance is ranked up).The interpretation of model by this model line
Sub-model provides (i.e. Logic Regression Models).
Specifically, threshold value mentioned above is by manually presetting, which is according to the quotient
Sample characteristics when family classification prediction model is trained are set, for example, when the sample characteristics are fraud trade company
Characteristic information when, the value of the preset threshold value can be any one value between greater than 0.6 and less than 0.9;
When the total probability corresponding to the trade company to be identified is greater than the preset threshold value, the type of the trade company to be identified
To cheat trade company;It is described to be identified when the total probability corresponding to the trade company to be identified is less than the preset threshold value
The type of trade company is normal trade company.The preset threshold value is embodied in the f () in Fig. 3 and between y, by adjusting
The tradeoff of recall rate and accuracy rate may be implemented in probability threshold value.According to the service condition of business side, need to some merchant type
Prediction recall rate and accuracy rate between do and weigh.In the scene for needing to identify fraud trade company, recall rate refers to model just
The fraud trade company Zhan really found always cheats the ratio of trade company, and the higher the better.What accuracy rate referred in model prediction result doubtful takes advantage of
It cheats and actually cheats trade company's ratio in trade company, the higher the better.Both but after the completion of model training, can not improve simultaneously.It mentions
High one will reduce another.When in business side, manpower is abundant, it is intended to improve recall rate.It can thus find as far as possible
More fraud trade companies.And when business side's manpower is in short supply, it is intended to improve accuracy rate, reduce the erroneous judgement to trade company.Model
Design is required to that the balance of the two is adjusted flexibly at any time according to business side's use demand.Such as the probability of setting suspect category
It is greater than a threshold value (such as 0.7) and just thinks that the trade company is suspicious trade company.By total probability corresponding to the merchant type with
Preset threshold value is compared, and determines the type of the trade company according to the result of the comparison.For example, enclosing training pattern
The training data of meal merchant type is separately input in 7 trained prediction submodels, obtains each enclosing merchant type of eating
Prediction probability;Obtain enclosing total training probability of meal merchant type after each training probability for enclosing meal merchant type is summarized;It will
The total training probability for enclosing meal merchant type is compared with preset threshold value, if enclosing total instruction of meal merchant type
It is big to practice the preset threshold value of likelihood ratio, then it is assumed that enclosing meal merchant type is Wei Can trade company, conversely, then enclosing meal user to be non-.
As one embodiment in the present invention, trade company can be the businessman using wechat payment gathering.Merchant type is
It is defined according to specific business scenario.Such as in fraud identification, normal trade company and fraud trade company can be divided into.Such as one
Business needs to identify Wei Can trade company.So we will remove one machine learning mould of training with the data of You Weican trade company sample
Type, so that model can recognize which trade company belongs to Wei Can trade company.If a business is only concerned whether trade company has solid shop/brick and mortar store
Transaction, we will be with there is the data of solid shop/brick and mortar store business sample to remove training machine learning model, so that model can recognize that pure line
Upper transaction trade company.It (has dinner as specific feature and concentrate on the specific period), is calculated automatically by the algorithm in model
It arrives.
The present invention pass through combine trade company to be identified transaction journal data and users personal data, by extract this two
The corresponding characteristic of class data, and be input in trained Merchant Category prediction model, it realizes to particular merchant class
Type is identified and is classified that the present invention can apply in related application, can be by precisely identifying trade company's class in wechat
Type, which precisely hits the illegal fraud on wechat payment platform, to have very great help, while the income for also protecting wechat to pay,
Preferential activity of the illegal trade company using wechat payment for specific type trade company is avoided to obtain illegitimate benefits.And the present invention
It is the type that can recognize trade company by simple pretreatment operation and Merchant Category prediction model, reduces nonterminal character and obtain institute
Need human cost, model deployment building complexity it is low, model it is versatile.
In one embodiment of the invention, a kind of merchant type identification device, the module architectures of described device are provided
Referring to Fig. 5, described device is comprised the following modules:
Request receiving module 510, for receiving merchant type identification request;
Data obtaining module 520, for obtaining the letter of trade company to be identified corresponding with merchant type identification request
Breath and userspersonal information;
Characteristic extracting module 530, for extract respectively the trade company to be identified information and the userspersonal information couple
The characteristic answered obtains the predictive data set of the trade company to be identified;
Data input module 540 predicts mould for the predictive data set to be input to trained Merchant Category
Type obtains the output result of the Merchant Category prediction model;
Trade company's identification module 550, for the output according to the Merchant Category prediction model as a result, determining described wait know
The type of other trade company.
Further, described device further includes training module 560, for training Merchant Category prediction model, the training
Module 560 includes:
Sample acquisition submodule 561, for obtaining information and the userspersonal information of sample trade company;
Sample characteristics extraction module 562, for extracting information and the userspersonal information couple of the sample trade company respectively
The sample characteristics data answered obtain the training dataset of the Merchant Category prediction model;
Model submodule 563, for each of the Merchant Category prediction model to be respectively trained in the training dataset
Submodel determines the parameter of each submodel of the Merchant Category prediction model, obtains the Merchant Category prediction model.
The sample acquisition submodule 561 includes:
Target trade company acquiring unit 5611, for obtaining the training samples information and related application of target trade company respectively
In all trade companies Transaction Information, the training samples information of the target trade company include the target trade company id information and
Corresponding remarks label, the remarks label are the corresponding merchant type of the target trade company manually marked, all quotient
The Transaction Information at family includes the id information and transaction journal information of all trade companies;
Matching unit 5612 is traversed, for traversing the Transaction Information of all trade companies, the ID of all trade companies is believed
Breath is matched with the id information of the target trade company, and the trade company of successful match in all trade companies is determined as sample quotient
Family, and using the remarks label as the merchant type label of the sample trade company;
User information acquiring unit 5613, for obtaining userspersonal information described in the related application.
The data input module 540 includes:
Probability calculation subelement 5411 predicts mould for the predictive data set to be separately input to the Merchant Category
Each submodel in type obtains sub- probability corresponding to the corresponding trade company to be identified of each submodel;
The total unit 5412 of probability calculation, for being carried out by aggregation function to probability corresponding to the trade company to be identified
It calculates, obtains total probability corresponding to the trade company to be identified;
Recognition unit 5413 determines the trade company to be identified for the total probability according to corresponding to the trade company to be identified
Type.
The recognition unit 5413 includes:
Comparing unit 54131, for carrying out total probability corresponding to the trade company to be identified and preset threshold value
Compare, determines the merchant type of the trade company to be identified according to the result of the comparison.
It should be understood that merchant type identification device provided by the above embodiment is when carrying out merchant type identification, only
The example of the division of the above functional modules, in practical application, can according to need and by above-mentioned function distribution by
Different functional modules is completed, i.e., the internal structure of merchant type identification device is divided into different functional modules, to complete
All or part of function described above.In addition, merchant type identification device embodiment provided in this embodiment and above-mentioned reality
Example offer trade company's kind identification method is provided and belongs to same design, specific implementation process is detailed in embodiment of the method, no longer superfluous here
It states.
Fig. 6 shows the flow diagram of trade company's prediction disaggregated model training and practice provided in an embodiment of the present invention.As
One embodiment of the present of invention, since the logic of identification merchant type is purely determined by business scenario, the present embodiment is for micro-
The complete procedure of training and practice that merchant type identification is illustrated for the identification of meal/Fei Weican trade company is enclosed in letter.This
The embodiment of the method that inventive embodiments provide can execute in mobile terminal, terminal or similar arithmetic unit,
In one embodiment of the invention, for running on the server.
Trade company predict disaggregated model training process the following steps are included:
Step 1: the background server of wechat application is from target trade company training sample database 61 and merchant information database
The training samples information of target trade company and the information of all trade companies in wechat application are obtained in 62 respectively.
Specifically, the training samples information of the target trade company includes the id information of the target trade company and is labeled as enclosing meal quotient
The remarks label of family type, the information of all trade companies in the wechat application of acquisition includes all trade company's id informations and friendship
Easy flowing water information.
Step 2: the information for all trade companies that server traversal is got, by the id information of all trade companies and target trade company
Id information matched, and the trade company of successful match in all trade companies is determined as sample trade company, and by the successful match
Trade company is labeled as the label of Wei Can trade company.
Step 3: server obtains the personal information of the wechat user in userspersonal information's database 63.
Step 4: the corresponding sample in the information for the sample trade company that server is got in extraction step 1 and step 2 respectively
The corresponding sample characteristics data of the userspersonal information got in eigen data and step 3, to realize 611 samples
Characteristic extracts corresponding process, obtains the set of the machine learning training data 612 of Merchant Category prediction model.
Step 5: server by training pattern 613, divides the set of the machine learning training data obtained in step 4
Each submodel of the Merchant Category prediction model is not trained, so that it is determined that each height of the Merchant Category prediction model
The parameter of model obtains Merchant Category prediction model 614, by the Merchant Category prediction model can to enclose meal/it is non-enclose meal quotient
Family is identified.
Trade company predict disaggregated model practice process the following steps are included:
Step 1: server obtains the information of trade company to be identified in merchant information database 62 and trade company's collective database 64
And in userspersonal information's database 63 wechat user personal information, prediction as merchant type to be identified it is original
Data source.
Step 2: server carries out the information of trade company to be identified and the predicted characteristics data of wechat userspersonal information respectively
621 are extracted, the set of the machine learning prediction data 622 of trade company to be identified is obtained.
The information of trade company to be identified and the corresponding characteristic of userspersonal information include but is not limited to:
1. transaction size: amount of money distribution, stroke count distribution, number of days, account number, payment failure rate etc.;
2. user characteristics: age, gender, repeat buying register source, using equipment, permanent provinces and cities etc.;
3. exchange hour: transaction accounting in different time periods;
4. paying denomination: the transaction accounting in different amount of money sections;Accounting highest transaction section;
5. the means of payment: channel, scene etc..
Step 3: 622 predictive data set of machine learning prediction data is separately input in Merchant Category prediction model 614
7 submodels, carry out model prediction 623,7 submodels and independently provide 7 groups of sub- probability corresponding to trade company to be identified,
This 7 groups of sub- probability are calculated by aggregation function, obtain total probability corresponding to trade company to be identified, it is true according to total probability
Fixed merchant type to be identified, obtained merchant type is Merchant Category prediction result 65.
It is illustrated in such a way that aggregation function is using hard ballot in the present embodiment.Since the target of model training is
Meal/Fei Weican trade company is enclosed in identification.3 are identified as enclosing meal in 7 submodels, and Wei Can trade company is 3 tickets, 4 be identified as it is non-enclose meal,
The non-meal that encloses is 4 tickets, so obtained total probability is to enclose meal to account for 3/7, non-to enclose meal to account for 4/7 final output merchant type be non-to enclose meal
Trade company.
The present invention passes through users personal data in the transaction journal data for combining trade company to be identified and wechat application, leads to
The corresponding characteristic for extracting these two types of data is crossed, and is input in trained Merchant Category prediction model, realization pair
Meal/non-meal merchant type of enclosing is enclosed to be identified and classified, it can be by precisely identifying that it is flat that merchant type precisely hits wechat payment
Illegal fraud on platform has very great help, while the income for also protecting wechat to pay, and illegal trade company is avoided to utilize wechat
Payment obtains illegitimate benefits for the preferential activity of specific type trade company.
Embodiment of the method provided in an embodiment of the present invention can be in mobile terminal, terminal or similar operation
It is executed in device, in one embodiment of the invention, for running on computer terminals, Fig. 7 is the embodiment of the present invention
Merchant type identification device terminal hardware block diagram.As shown in fig. 7, terminal 800 may include RF
(Radio Frequency, radio frequency) circuit 110, the memory for including one or more computer readable storage medium
120, input unit 130, display unit 140, sensor 150, voicefrequency circuit 160, WiFi (wireless fidelity, nothing
Line fidelity) module 170, the processor 180 and the components such as power supply 190 that include one or more than one processing core.
It will be understood by those skilled in the art that the restriction of the not structure paired terminal of terminal structure shown in Fig. 7, may include than figure
Show more or fewer components, perhaps combines certain components or different component layouts.Wherein:
RF circuit 110 can be used for receiving and sending messages or communication process in, signal sends and receivees, particularly, by base station
After downlink information receives, one or the processing of more than one processor 180 are transferred to;In addition, the data for being related to uplink are sent to
Base station.In general, RF circuit 110 include but is not limited to antenna, at least one amplifier, tuner, one or more oscillator,
It is subscriber identity module (SIM) card, transceiver, coupler, LNA (Low Noise Amplifier, low-noise amplifier), double
Work device etc..In addition, RF circuit 110 can also be communicated with network and other equipment by wireless communication.The wireless communication can be with
Use any communication standard or agreement, including but not limited to GSM (Global System of Mobile
Communication, global system for mobile communications), GPRS (General Packet Radio Service, general packet without
Line service), CDMA (Code Division Multiple Access, CDMA), WCDMA (Wideband Code
Division Multiple Access, wideband code division multiple access), LTE (Long Term Evolution, long term evolution), electricity
Sub- mail, SMS (Short Messaging Service, short message service) etc..
Memory 120 can be used for storing software program and module, and processor 180 is stored in memory 120 by operation
Software program and module, thereby executing various function application and data processing.Memory 120 can mainly include storage
Program area and storage data area, wherein storing program area can (such as the sound of application program needed for storage program area, function
Playing function, image player function etc.) etc.;Storage data area, which can be stored, uses created data (ratio according to terminal 800
Such as audio data, phone directory) etc..In addition, memory 120 may include high-speed random access memory, it can also include non-
Volatile memory, for example, at least a disk memory, flush memory device or other volatile solid-state parts.Accordingly
Ground, memory 120 can also include Memory Controller, to provide processor 180 and input unit 130 to memory 120
Access.
Input unit 130 can be used for receiving the number or character information of input, and generate and user setting and function
Control related keyboard, mouse, operating stick, optics or trackball signal input.Specifically, input unit 130 may include touching
Sensitive surfaces 131 and other input equipments 132.Touch sensitive surface 131, also referred to as touch display screen or Trackpad are collected and are used
Family on it or nearby touch operation (such as user using any suitable object or attachment such as finger, stylus in touch-sensitive table
Operation on face 131 or near touch sensitive surface 131), and corresponding attachment device is driven according to preset formula.It can
Choosing, touch sensitive surface 131 may include both touch detecting apparatus and touch controller.Wherein, touch detecting apparatus detects
The touch orientation of user, and touch operation bring signal is detected, transmit a signal to touch controller;Touch controller from
Touch information is received on touch detecting apparatus, and is converted into contact coordinate, then gives processor 180, and can reception processing
Order that device 180 is sent simultaneously is executed.Furthermore, it is possible to more using resistance-type, condenser type, infrared ray and surface acoustic wave etc.
Seed type realizes touch sensitive surface 131.In addition to touch sensitive surface 131, input unit 130 can also include other input equipments 132.
Specifically, other input equipments 132 can include but is not limited to physical keyboard, function key (such as volume control button, switch
Key etc.), trace ball, mouse, one of operating stick etc. or a variety of.
Display unit 140 can be used for showing information input by user or the information and terminal 800 that are supplied to user
Various graphical user interface, these graphical user interface can be by figure, text, icon, video and any combination thereof come structure
At.Display unit 140 may include display panel 141, optionally, can use LCD (Liquid Crystal Display, liquid
Crystal display), the forms such as OLED (Organic Light-Emitting Diode, Organic Light Emitting Diode) configure display
Panel 141.Further, touch sensitive surface 131 can cover display panel 141, when touch sensitive surface 131 detect it is on it or attached
After close touch operation, processor 180 is sent to determine the type of touch event, is followed by subsequent processing device 180 according to touch event
Type corresponding visual output is provided on display panel 141.Although in Fig. 7, touch sensitive surface 131 and display panel 141
It is that input and input function are realized as two independent components, but in some embodiments it is possible to by touch sensitive surface
131 integrate with display panel 141 and realize and output and input function.
Terminal 800 may also include at least one sensor 150, such as optical sensor, motion sensor and other sensings
Device.Specifically, optical sensor may include ambient light sensor and proximity sensor, wherein ambient light sensor can be according to environment
The light and shade of light adjusts the brightness of display panel 141, and proximity sensor can close display when terminal 800 is moved in one's ear
Panel 141 and/or backlight.As a kind of motion sensor, gravity accelerometer can detect in all directions (general
For three axis) size of acceleration, it can detect that size and the direction of gravity when static, can be used to identify the application of terminal posture
(such as horizontal/vertical screen switching, dependent game, magnetometer pose calibrating), Vibration identification correlation function (such as pedometer, percussion)
Deng;Other sensings such as gyroscope, barometer, hygrometer, thermometer, infrared sensor for can also configure as terminal 800
Device, details are not described herein.
Voicefrequency circuit 160, loudspeaker 161, microphone 162 can provide the audio interface between user and terminal 800.Sound
Electric signal after the audio data received conversion can be transferred to loudspeaker 161, be converted by loudspeaker 161 by frequency circuit 160
For voice signal output;On the other hand, the voice signal of collection is converted to electric signal by microphone 162, by voicefrequency circuit 160
Audio data is converted to after reception, then by after the processing of audio data output processor 180, through RF circuit 110 to be sent to such as
Another terminal, or audio data is exported to memory 120 to be further processed.Voicefrequency circuit 160 is also possible that ear
Tip jack, to provide the communication of peripheral hardware earphone Yu terminal 800.
WiFi belongs to short range wireless transmission technology, and terminal 800 can help user to receive and dispatch electricity by WiFi module 170
Sub- mail, browsing webpage and access streaming video etc., it provides wireless broadband internet access for user.Although Fig. 7 shows
Go out WiFi module 170, but it is understood that, and it is not belonging to must be configured into for terminal 800, it completely can be according to need
It to omit within the scope of not changing the essence of the invention.
Processor 180 is the control centre of terminal 800, utilizes each portion of various interfaces and the entire terminal of connection
Point, by running or execute the software program and/or module that are stored in memory 120, and calls and be stored in memory
Data in 120 execute the various functions and processing data of terminal 800, to carry out integral monitoring to terminal.Optionally, locate
Managing device 180 may include one or more processing cores;Preferably, processor 180 can integrate application processor and modulatedemodulate is mediated
Manage device, wherein the main processing operation system of application processor, user interface and application program etc., modem processor is main
Processing wireless communication.It is understood that above-mentioned modem processor can not also be integrated into processor 180.
Terminal 800 further includes the power supply 190 (such as battery) powered to all parts, it is preferred that power supply can pass through electricity
Management system and processor 180 are logically contiguous, to realize management charging, electric discharge and power consumption by power-supply management system
The functions such as management.Power supply 190 can also include one or more direct current or AC power source, recharging system, power supply event
Hinder the random components such as detection circuit, power adapter or inverter, power supply status indicator.
Although being not shown, terminal 800 can also include camera, bluetooth module etc., and details are not described herein.Specifically at this
In embodiment, the display unit of terminal is touch-screen display, terminal further include have memory and one or one with
On program, one of them perhaps more than one program be stored in memory and be configured to by one or more than one
Processor execution states one or more than one program includes the instruction for performing the following operation:
Receive merchant type identification request;
Obtain information and the userspersonal information of trade company to be identified corresponding with merchant type identification request;
The information and the corresponding characteristic of the userspersonal information for extracting the trade company to be identified respectively, obtain institute
State the predictive data set of trade company to be identified;
The predictive data set is input to trained Merchant Category prediction model, it is pre- to obtain the Merchant Category
Survey the output result of model;And
According to the output of the Merchant Category prediction model as a result, determining the type of the trade company to be identified.
Specifically, the processor of terminal is also used to execute the instruction operated below: the training Merchant Category prediction mould
Type, which comprises
Obtain information and the userspersonal information of sample trade company;
The information and the corresponding sample characteristics data of userspersonal information for extracting the sample trade company respectively, obtain institute
State the training dataset of Merchant Category prediction model;
The training dataset is respectively trained to each submodel of the Merchant Category prediction model, determines the quotient
The parameter of each submodel of family classification prediction model, obtains the Merchant Category prediction model.
Specifically, the processor of terminal is also used to execute the instruction operated below:
The training samples information of target trade company and the information of all trade companies in related application, the mesh are obtained respectively
The training samples information of mark trade company includes the id information and corresponding remarks label of the target trade company, and the remarks label is
The corresponding merchant type of the target trade company manually marked, the information of all trade companies include the ID of all trade companies
Information and transaction journal information;
The information for traversing all trade companies, by the ID information of the id information of all trade companies and the target trade company
It is matched, the trade company of successful match in all trade companies is determined as sample trade company, and using the remarks label as institute
State the merchant type label of sample trade company;
Obtain userspersonal information described in the related application.
Specifically, the processor of terminal is also used to execute the instruction operated below:
The predictive data set is separately input to each submodel in the Merchant Category prediction model, obtains institute
State sub- probability corresponding to the corresponding trade company to be identified of each submodel;
Sub- probability corresponding to the trade company to be identified is calculated by aggregation function, obtains the quotient to be identified
Total probability corresponding to family;
According to total probability corresponding to the trade company to be identified, the type of the trade company to be identified is determined.
Specifically, the processor of terminal is also used to execute the instruction operated below: will be corresponding to the trade company to be identified
Total probability is compared with preset threshold value, determines the merchant type of the trade company to be identified according to the result of the comparison.
In one embodiment of the invention, a kind of computer readable storage medium is provided, the computer-readable storage
Medium can be computer readable storage medium included in the memory in above-described embodiment;It is also possible to individualism,
Without the computer readable storage medium in supplying terminal.Computer-readable recording medium storage have one or more than one
Program, one or more than one program are used to execute merchant type recognition methods by one or more than one processor
Instruction, the method instruction includes:
Receive merchant type identification request;
Obtain information and the userspersonal information of trade company to be identified corresponding with merchant type identification request;
The information and the corresponding characteristic of the userspersonal information for extracting the trade company to be identified respectively, obtain institute
State the predictive data set of trade company to be identified;
The predictive data set is input to trained Merchant Category prediction model, it is pre- to obtain the Merchant Category
Survey the output result of model;And
According to the output of the Merchant Category prediction model as a result, determining the type of the trade company to be identified.
Further, the method includes the training Merchant Category prediction model, the training process includes:
Obtain information and the userspersonal information of sample trade company;
The information and the corresponding sample characteristics data of userspersonal information for extracting the sample trade company respectively, obtain institute
State the training dataset of Merchant Category prediction model;
The training dataset is respectively trained to each submodel of the Merchant Category prediction model, determines the quotient
The parameter of each submodel of family classification prediction model, obtains the Merchant Category prediction model.
Further, the information for obtaining sample trade company and userspersonal information, comprising:
The training samples information of target trade company and the information of all trade companies in related application, the mesh are obtained respectively
The training samples information of mark trade company includes the id information and corresponding remarks label of the target trade company, and the remarks label is
The corresponding merchant type of the target trade company manually marked, the information of all trade companies include the ID of all trade companies
Information and transaction journal information;
The information for traversing all trade companies, by the ID information of the id information of all trade companies and the target trade company
It is matched, the trade company of successful match in all trade companies is determined as sample trade company, and using the remarks label as institute
State the merchant type label of sample trade company;
Obtain userspersonal information described in the related application.
Further, described that the predictive data set is input to trained Merchant Category prediction model, obtain institute
State the output result of Merchant Category prediction model, comprising:
The predictive data set is separately input to each submodel in the Merchant Category prediction model, obtains institute
State sub- probability corresponding to the corresponding trade company to be identified of each submodel;
Sub- probability corresponding to the trade company to be identified is calculated by aggregation function, obtains the quotient to be identified
Total probability corresponding to family;
According to total probability corresponding to the trade company to be identified, the type of the trade company to be identified is determined.
Further, the total probability according to corresponding to the trade company to be identified, determines the type of the trade company, packet
It includes: total probability corresponding to the trade company to be identified being compared with preset threshold value, is determined according to the result of the comparison
The merchant type of the trade company to be identified.
The serial number of the above embodiments of the invention is only for description, does not represent the advantages or disadvantages of the embodiments.
Those of ordinary skill in the art will appreciate that realizing that all or part of the steps of above-described embodiment can pass through hardware
It completes, relevant hardware can also be instructed to complete by program, the program can store in a kind of computer-readable
In storage medium, storage medium mentioned above can be read-only memory, disk or CD etc..
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (11)
1. a kind of merchant type recognition methods, which is characterized in that the described method includes:
Receive merchant type identification request;
Obtain information and the userspersonal information of trade company to be identified corresponding with merchant type identification request;
The information and the corresponding characteristic of the userspersonal information for extracting the trade company to be identified respectively obtain described wait know
The predictive data set of other trade company;
The predictive data set is input to trained Merchant Category prediction model, obtains the Merchant Category prediction model
Output result;And
According to the output of the Merchant Category prediction model as a result, determining the type of the trade company to be identified.
2. the method according to claim 1, wherein the method also includes the training Merchant Categories to predict mould
Type, comprising:
Obtain information and the userspersonal information of sample trade company;
The information and the corresponding sample characteristics data of userspersonal information for extracting the sample trade company respectively, obtain the trade company
The training dataset of classification prediction model;
The training dataset is respectively trained to each submodel of the Merchant Category prediction model, determines the Merchant Category
The parameter of each submodel of prediction model obtains the Merchant Category prediction model.
3. according to the method described in claim 2, it is characterized in that, the information for obtaining sample trade company and individual subscriber letter
Breath, comprising:
The training samples information of target trade company and the information of all trade companies in related application, the target trade company are obtained respectively
Training samples information include the target trade company id information and corresponding remarks label, the remarks label is artificial mark
The corresponding merchant type of the target trade company, the information of all trade companies includes id information and the friendship of all trade companies
Easy flowing water information;
The information for traversing all trade companies, by the id information of all trade companies and the progress of the id information of the target trade company
Match, the trade company of successful match in all trade companies is determined as sample trade company, and using the remarks label as the sample
The merchant type label of trade company;
Obtain userspersonal information described in the related application.
4. the method according to claim 1, wherein it is described the predictive data set is input to it is trained
Merchant Category prediction model obtains the output result of the Merchant Category prediction model, comprising:
The predictive data set is separately input to each submodel in the Merchant Category prediction model, is obtained described each
Sub- probability corresponding to the corresponding trade company to be identified of submodel;
Sub- probability corresponding to the trade company to be identified is calculated by aggregation function, it is right to obtain the institute, trade company to be identified
The total probability answered;
According to total probability corresponding to the trade company to be identified, the type of the trade company to be identified is determined.
5. according to the method described in claim 4, it is characterized in that, described total general according to corresponding to the trade company to be identified
Rate determines the type of the trade company, comprising:
Total probability corresponding to the trade company to be identified is compared with preset threshold value, is determined according to the result of the comparison
The merchant type of the trade company to be identified.
6. according to the method described in claim 5, it is characterized in that, which comprises
When the sample characteristics are the characteristic information for cheating trade company, the value of the preset threshold value is between 0.6 to 1
When any one value;
If total probability corresponding to the trade company to be identified is greater than the preset threshold value, the type of the trade company to be identified
To cheat trade company;
If total probability corresponding to the trade company to be identified is less than the preset threshold value, the type of the trade company to be identified
For normal trade company.
7. a kind of merchant type identification device, which is characterized in that described device includes:
Request receiving module, for receiving merchant type identification request;
Data obtaining module, for obtaining the information and use of trade company to be identified corresponding with merchant type identification request
Family personal information;
Characteristic extracting module, for extracting the information and the corresponding feature of the userspersonal information of the trade company to be identified respectively
Data obtain the predictive data set of the trade company to be identified;
Data input module obtains institute for the predictive data set to be input to trained Merchant Category prediction model
State the output result of Merchant Category prediction model;
Trade company's identification module, for the output according to the Merchant Category prediction model as a result, determining the trade company to be identified
Type.
8. device according to claim 7, which is characterized in that further include training module, for training Merchant Category to predict
Model, the training module include:
Sample acquisition submodule, for obtaining information and the userspersonal information of sample trade company;
Sample characteristics extraction module, for extracting the information and the corresponding sample of userspersonal information of the sample trade company respectively
Characteristic obtains the training dataset of the Merchant Category prediction model;
Model submodule, for the training dataset to be respectively trained to each submodel of the Merchant Category prediction model,
The parameter for determining each submodel of the Merchant Category prediction model obtains the Merchant Category prediction model.
9. device according to claim 8, which is characterized in that the sample acquisition submodule includes:
Target trade company acquiring unit, all quotient in training samples information and related application for obtaining target trade company respectively
The information at family, the training samples information of the target trade company include the id information and corresponding remarks label of the target trade company,
The remarks label is the corresponding merchant type of the target trade company manually marked, and the information of all trade companies includes described
The id information and transaction journal information of all trade companies;
Matching unit is traversed, for traversing the information of all trade companies, by the id information of all trade companies and the target
The id information of trade company is matched, and the trade company of successful match in all trade companies is determined as sample trade company, and will be described standby
Infuse merchant type label of the label as the sample trade company;
User information acquiring unit, for obtaining userspersonal information described in the related application.
10. device according to claim 7, which is characterized in that the data input module includes:
Probability calculation subelement, it is each in the Merchant Category prediction model for the predictive data set to be separately input to
Submodel obtains sub- probability corresponding to the corresponding trade company to be identified of each submodel;
The total unit of probability calculation is obtained for being calculated by aggregation function sub- probability corresponding to the trade company to be identified
To total probability corresponding to the trade company to be identified;
Recognition unit determines the type of the trade company to be identified for the total probability according to corresponding to the trade company to be identified.
11. device according to claim 7, which is characterized in that the recognition unit includes:
Comparing unit, for total probability corresponding to the trade company to be identified to be compared with preset threshold value, according to
Comparison result determines the merchant type of the trade company to be identified.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711416934.3A CN109961296A (en) | 2017-12-25 | 2017-12-25 | Merchant type recognition methods and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711416934.3A CN109961296A (en) | 2017-12-25 | 2017-12-25 | Merchant type recognition methods and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109961296A true CN109961296A (en) | 2019-07-02 |
Family
ID=67020550
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711416934.3A Pending CN109961296A (en) | 2017-12-25 | 2017-12-25 | Merchant type recognition methods and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109961296A (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110458581A (en) * | 2019-07-11 | 2019-11-15 | 阿里巴巴集团控股有限公司 | Merchant business turnover abnormal recognition methods and device |
CN110610575A (en) * | 2019-09-20 | 2019-12-24 | 北京百度网讯科技有限公司 | Coin identification method and device and cash register |
CN110619585A (en) * | 2019-08-16 | 2019-12-27 | 广州越秀金融科技有限公司 | Method, device, storage medium and processor for recommending data |
CN110750697A (en) * | 2019-10-30 | 2020-02-04 | 汉海信息技术(上海)有限公司 | Merchant classification method, device, equipment and storage medium |
CN111159398A (en) * | 2019-12-05 | 2020-05-15 | 银联智策顾问(上海)有限公司 | Method and device for identifying merchant type |
CN111428217A (en) * | 2020-04-12 | 2020-07-17 | 中信银行股份有限公司 | Method and device for identifying cheat group, electronic equipment and computer readable storage medium |
CN111784448A (en) * | 2020-06-24 | 2020-10-16 | 支付宝(杭州)信息技术有限公司 | Merchant data processing method and system |
CN112581271A (en) * | 2020-12-21 | 2021-03-30 | 上海浦东发展银行股份有限公司 | Merchant transaction risk monitoring method, device, equipment and storage medium |
CN112732912A (en) * | 2020-12-30 | 2021-04-30 | 平安科技(深圳)有限公司 | Sensitive tendency expression detection method, device, equipment and storage medium |
WO2021164232A1 (en) * | 2020-02-17 | 2021-08-26 | 平安科技(深圳)有限公司 | User identification method and apparatus, and device and storage medium |
CN113554099A (en) * | 2021-07-27 | 2021-10-26 | 中国银联股份有限公司 | Method and device for identifying abnormal commercial tenant |
CN113592160A (en) * | 2021-07-13 | 2021-11-02 | 北京淇瑀信息科技有限公司 | Fusion processing-based loss of contact risk prediction method and system for user equipment and computer equipment |
CN113971863A (en) * | 2021-10-19 | 2022-01-25 | 无锡格林通安全装备有限公司 | Flame detection method based on FFT-GMM-K neighbor |
WO2023244164A1 (en) * | 2022-06-13 | 2023-12-21 | Gp Network Asia Pte. Ltd. | Method and device for detecting merchant from payment transaction |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105184574A (en) * | 2015-06-30 | 2015-12-23 | 电子科技大学 | Method for detecting fraud behavior of merchant category code cloning |
CN105590261A (en) * | 2014-12-31 | 2016-05-18 | 中国银联股份有限公司 | Merchant risk estimation method and system |
WO2017041651A1 (en) * | 2015-09-09 | 2017-03-16 | 阿里巴巴集团控股有限公司 | User data classification method and device |
CN106650721A (en) * | 2016-12-28 | 2017-05-10 | 吴晓军 | Industrial character identification method based on convolution neural network |
CN107067293A (en) * | 2017-03-07 | 2017-08-18 | 北京三快在线科技有限公司 | Merchant category method, device and electronic equipment |
CN107301221A (en) * | 2017-06-16 | 2017-10-27 | 华南理工大学 | A kind of data digging method of multiple features dimension heap fusion |
CN107465741A (en) * | 2017-08-02 | 2017-12-12 | 北京小度信息科技有限公司 | Information-pushing method and device |
-
2017
- 2017-12-25 CN CN201711416934.3A patent/CN109961296A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105590261A (en) * | 2014-12-31 | 2016-05-18 | 中国银联股份有限公司 | Merchant risk estimation method and system |
CN105184574A (en) * | 2015-06-30 | 2015-12-23 | 电子科技大学 | Method for detecting fraud behavior of merchant category code cloning |
WO2017041651A1 (en) * | 2015-09-09 | 2017-03-16 | 阿里巴巴集团控股有限公司 | User data classification method and device |
CN106650721A (en) * | 2016-12-28 | 2017-05-10 | 吴晓军 | Industrial character identification method based on convolution neural network |
CN107067293A (en) * | 2017-03-07 | 2017-08-18 | 北京三快在线科技有限公司 | Merchant category method, device and electronic equipment |
CN107301221A (en) * | 2017-06-16 | 2017-10-27 | 华南理工大学 | A kind of data digging method of multiple features dimension heap fusion |
CN107465741A (en) * | 2017-08-02 | 2017-12-12 | 北京小度信息科技有限公司 | Information-pushing method and device |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110458581A (en) * | 2019-07-11 | 2019-11-15 | 阿里巴巴集团控股有限公司 | Merchant business turnover abnormal recognition methods and device |
CN110458581B (en) * | 2019-07-11 | 2024-01-16 | 创新先进技术有限公司 | Method and device for identifying business turnover abnormality of commercial tenant |
CN110619585A (en) * | 2019-08-16 | 2019-12-27 | 广州越秀金融科技有限公司 | Method, device, storage medium and processor for recommending data |
CN110610575A (en) * | 2019-09-20 | 2019-12-24 | 北京百度网讯科技有限公司 | Coin identification method and device and cash register |
US11354887B2 (en) | 2019-09-20 | 2022-06-07 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Coin identification method, device, and cash register |
CN110750697A (en) * | 2019-10-30 | 2020-02-04 | 汉海信息技术(上海)有限公司 | Merchant classification method, device, equipment and storage medium |
CN111159398A (en) * | 2019-12-05 | 2020-05-15 | 银联智策顾问(上海)有限公司 | Method and device for identifying merchant type |
CN111159398B (en) * | 2019-12-05 | 2023-10-31 | 银联智策顾问(上海)有限公司 | Method and device for identifying merchant types |
WO2021164232A1 (en) * | 2020-02-17 | 2021-08-26 | 平安科技(深圳)有限公司 | User identification method and apparatus, and device and storage medium |
CN111428217A (en) * | 2020-04-12 | 2020-07-17 | 中信银行股份有限公司 | Method and device for identifying cheat group, electronic equipment and computer readable storage medium |
CN111784448A (en) * | 2020-06-24 | 2020-10-16 | 支付宝(杭州)信息技术有限公司 | Merchant data processing method and system |
CN112581271A (en) * | 2020-12-21 | 2021-03-30 | 上海浦东发展银行股份有限公司 | Merchant transaction risk monitoring method, device, equipment and storage medium |
CN112581271B (en) * | 2020-12-21 | 2022-11-15 | 上海浦东发展银行股份有限公司 | Merchant transaction risk monitoring method, device, equipment and storage medium |
CN112732912A (en) * | 2020-12-30 | 2021-04-30 | 平安科技(深圳)有限公司 | Sensitive tendency expression detection method, device, equipment and storage medium |
CN112732912B (en) * | 2020-12-30 | 2024-04-09 | 平安科技(深圳)有限公司 | Sensitive trend expression detection method, device, equipment and storage medium |
CN113592160A (en) * | 2021-07-13 | 2021-11-02 | 北京淇瑀信息科技有限公司 | Fusion processing-based loss of contact risk prediction method and system for user equipment and computer equipment |
CN113554099A (en) * | 2021-07-27 | 2021-10-26 | 中国银联股份有限公司 | Method and device for identifying abnormal commercial tenant |
CN113971863A (en) * | 2021-10-19 | 2022-01-25 | 无锡格林通安全装备有限公司 | Flame detection method based on FFT-GMM-K neighbor |
WO2023244164A1 (en) * | 2022-06-13 | 2023-12-21 | Gp Network Asia Pte. Ltd. | Method and device for detecting merchant from payment transaction |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109961296A (en) | Merchant type recognition methods and device | |
CN106446797B (en) | Image clustering method and device | |
CN108875781A (en) | A kind of labeling method, apparatus, electronic equipment and storage medium | |
CN104618415B (en) | Credit accounts creation method, apparatus and system | |
CN107944380A (en) | Personal identification method, device and storage device | |
CN108470253B (en) | User identification method and device and storage equipment | |
CN104392350B (en) | A kind of processing method, the device and system of virtual resource data | |
CN104951960B (en) | Exchange method and device based on object message | |
CN107992728A (en) | Face verification method and device | |
CN108121803A (en) | A kind of method and server of definite page layout | |
CN107368872B (en) | A kind of method and apparatus generating bar code image | |
CN110516967A (en) | A kind of method and relevant apparatus of information evaluation | |
CN104160734A (en) | Measuring quality of experience associated with a mobile device | |
CN110503409B (en) | Information processing method and related device | |
CN103714316B (en) | Image-recognizing method, device and electronic equipment | |
CN109905545A (en) | A kind of message treatment method and terminal, computer readable storage medium | |
CN109256147A (en) | Audio cadence detection method, device and storage medium | |
CN107783906A (en) | Method of testing, device and storage medium | |
CN109194689A (en) | Abnormal behaviour recognition methods, device, server and storage medium | |
CN113596844A (en) | Early warning method, device, medium and electronic equipment based on data information | |
CN109544139A (en) | A kind of method of payment, terminal and readable storage medium storing program for executing | |
CN109992965A (en) | Process handling method and device, electronic equipment, computer readable storage medium | |
CN111708944A (en) | Multimedia resource identification method, device, equipment and storage medium | |
US20170034092A1 (en) | Method and device for processing short message | |
CN108763509A (en) | A kind of data processing method, mobile terminal |
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 |