Specific embodiment
In order to make those skilled in the art better understand the technical solutions in the application, below in conjunction with the application reality
The attached drawing in example is applied, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described implementation
Example is merely a part but not all of the embodiments of the present application.Based on the embodiment in the application, this field is common
The application protection all should belong in technical staff's every other embodiment obtained without creative efforts
Range.
Fig. 1 shows the signal of the application scenarios of the exchange information processing method provided according to some embodiments of the present application
Block diagram.Shown in referring to Fig.1, which may include: at least one client 110 and server end 120.Client
It is communicated between 110 and server end 120 by network 130.Bank or Third-party payment mechanism are installed in client 110
Payment application, user can be carried out under line by the payment application in client 110 or the trading activity that pays on line, such as
The trading activities such as two dimensional code is paid, small amount exempts from close payment, transfers accounts or supplements with money are swept, and record friendship related with corresponding trading activity
Easy behavioral data.The trading activity that server end 120 executes client 110 confirms, completes corresponding trading activity, and
Record trading activity data related with the trading activity.
It should be noted that client 110 can calculate for mobile phone, tablet computer, desktop computer, Portable notebook type
Machine or POS (Point Of Sales, point of sale) terminal etc..Server 120 can be the physical server comprising unique host,
Perhaps for mainframe cluster carrying virtual server or be Cloud Server.Network 130 can be cable network or wireless network
Network, for example, network 130 can be Public Switched Telephone Network (Public Switched Telephone Network, PSTN)
Or internet.
It should be noted that step in exchange information processing method in the example embodiment of the application can part by
Client 110 execute, partially executed, can also all be executed by server 120 by server 120, the present invention to this without
Particular determination.
Below with reference to the application scenarios of Fig. 1, it is described with reference to Figure 2 the Transaction Information of the exemplary embodiment according to the application
Processing method.It should be noted which is shown only for the purpose of facilitating an understanding of the spirit and principles of the present invention for above-mentioned application scenarios, this
The embodiment of invention is unrestricted in this regard.On the contrary, the embodiment of the present invention can be applied to applicable any scene.
Fig. 2 shows the flow diagrams of the exchange information processing method provided according to some embodiments of the present application, should
Exchange information processing method can be applied to the server 120 in Fig. 1.Referring to shown in Fig. 2, which includes
Step S210 to step S240, is below described in detail the exchange information processing method in the example embodiment of Fig. 2.
Referring to shown in Fig. 2, in step S210, the trading activity data based on target subject generate the row of the target subject
For sequence.
In the exemplary embodiment, target subject is the main body for being able to reflect the attaching information for the user for carrying out trading activity,
The attaching information can be to carry out geographical location information locating for the user of trading activity, which can be to go with transaction
For corresponding bank account information such as bank card number and bank card number, IP (Internet Protocol, network association
View) address, telephone number, identification card number information, or other main bodys appropriate equipment for example corresponding with trading activity
IMEI (International Mobile Equipment Identity, International Mobile Equipment Identity code), equipment MAC
The main bodys such as (Media Access Control, media access control) address, this is equally within the scope of protection of this application.Below
It is illustrated by taking bank card number as an example in section Example mode.
In the exemplary embodiment, the trading activity data of available target subject within a predetermined period of time, the pre- timing
Between section can be the periods such as 15 days, 1 month, 2 months or 3 months.The trading activity data of target subject may include target
Transfer information, payment information and the Receiving information of main body also may include that other data related with trading activity are for example stepped on
Data, load value data or purchase data are recorded, the application is to this without particular determination.
Further, the behavior of the target subject is generated based on the trading activity data of target subject within a predetermined period of time
Sequence, the behavior sequence of target subject are the transaction row according to the target subject in some time section of transaction time of origin sequence
For data, for example, the trading activity data set based on target subject in predetermined amount of time such as 1 month generate the target subject
Behavior sequence, then the trading activity data in the predetermined amount of time can be divided based on predetermined space such as 1 day, be obtained
To the trading activity data in some time section, and according to time of origin by the trading activity data in each time interval into
Row sequence, generates the behavior sequence in the predetermined amount of time corresponding with the target subject.
In step S220, the trading activity feature of the target subject is extracted from the behavior sequence of the target subject,
In, which includes the main body variable information of target subject corresponding with trading activity data, main body variable letter
The information of attaching information of the breath comprising indicating the target subject.
In the exemplary embodiment, the trading activity feature of target subject may include target subject main body variable information,
Transaction amount, beneficiary information or paying party information, loco, exchange hour and type of transaction feature also may include
The characteristic informations such as other trading activity features appropriate such as means of payment, coupon use information.The main body of target subject becomes
Measure information include indicate target subject attaching information information, such as target subject be bank card when, the main body of target subject
Variable information includes bank's card number, bank card BIN (Bank Identification Number, the bank that bank's card number includes
Identification code) it can indicate the attaching information of bank card, bank card BIN is usually indicated by 6 bit digitals, appears in bank card number
First 6, such as the card BIN of business bank, Hefei City is 603601, the card BIN of Wuxi City business bank is 603265, Changshu City
The card BIN of Rural Commercial Bank is 603694, the card BIN of business bank, Daliang City is 603708.It is identity card in target subject
When, the main body variable information of target subject includes identification card number, and first 6 of identification card number can indicate the ownership letter of identity card
Breath, such as Beijing are 110000, Tianjin 120000, Hebei province 130000, Shanxi Province 140000, and the Inner Mongol is autonomous
Area is 150000.
Further, in the exemplary embodiment, the trading activity data in the behavior sequence of target subject are segmented
Processing extracts from the trading activity data in the behavior sequence after participle and identifies related word such as target master with transaction risk
The information such as the main body variable of body, transaction amount, type of transaction, exchange hour, to extract with transaction risk identify related word into
Row coding such as one-hot one-hot coding, generate the feature of corresponding with the behavior sequence of target subject trading activity feature to
Amount.For example, when identifying that related word includes the main body variable of target subject, transaction amount, type of transaction with transaction risk, mesh
The one-hot coding for marking the main body variable of main body is [1,0,0,0], and the one-hot coding of transaction amount is [0,1,0,0], type of transaction
One-hot coding be [0,0,1,0], based on identifying that the coding generation of related word is corresponding in trading activity data with transaction risk
Trading activity feature feature vector.
It should be noted that, although trading activity data are encoded using one-hot coding in this exemplary embodiment,
But in the example embodiment of the application, trading activity data can also be encoded using other coding modes appropriate
Such as bag of words, TF-IDF (Term Frequency-Inverse Document Frequency, term frequency-inverse document frequency
Rate) model etc., the application is to this without particular determination.
Next, in step S230, the dimensional information of the trading activity feature based on target subject is embedded in by word
Mode generates insertion vector corresponding with the trading activity feature.
In the exemplary embodiment, the dimensional information of the trading activity feature of target subject is special for identifying the trading activity
The same or similar characteristic information in sign, the dimensional information of the trading activity feature of target subject may include: that trading activity is returned
Belong to one of dimensional information, trading activity type dimension information, transaction amount dimensional information and exchange hour dimensional information
Or it is a variety of.For example, trading activity ownership dimensional information may include bank card number information, IP address information, cell-phone number information,
Identification card number information etc. belongs to dimensional information;Trading activity type dimension information may include: consumer transaction, insurance class friendship
Easily, financing class trades, borrows the type dimensions information such as the transaction of refund class, the transaction of wage class;Transaction amount dimensional information may include
Multiple amount of money sections such as amount of money dimension such as [0,5000], [5000,10000], [10000,20000], [20000,50000] is believed
Breath;Exchange hour dimensional information may include that multiple time intervals set predetermined amount of time for example as 2018.1.1 to 2018.5.1,
Then multiple time intervals can for [2018.1.1,2018.2.1], [2018.2.1,2018.3.1], [2018.3.1,
2018.4.1], the time dimensions information such as [2018.4.1,2018.5.1].
Further, the dimensional information based on trading activity feature generates and the trading activity in such a way that word is embedded in
The corresponding insertion vector of feature.Word insertion is that a kind of distribution of word indicates, can be by a high dimension in such a way that word is embedded in
Word space reflection into the vector space of a low-dimensional number.In the exemplary embodiment, the dimension of trading activity feature is being determined
It spends after information, word incorporation model such as LSTM (Long Short-Term Memory, shot and long term memory) model can be passed through
By the trading activity Feature Mapping of multidimensional to the insertion vector of low-dimensional number corresponding with the dimensional information, calculated so as to reduce
Amount improves data-handling efficiency.Such as, it is determined that the dimensional information of trading activity feature includes the information of four dimensions, that is, is handed over
Easy is to belong to dimensional information, trading activity type dimension information, transaction amount dimensional information and exchange hour dimensional information,
It can be by word incorporation model by the insertion vector of trading activity Feature Mapping to the four dimensions.
It should be noted that in this exemplary embodiment, the mode of word insertion can be with LSTM model, or
Word2Vec model or Glove model, can also for other neural network models appropriate such as Skip-Gram model or
The models such as CBOW (Continuous Bag-of-Words, continuous bag of words), this is equally within the scope of protection of this application.
In step S240, the main body variable information based on target subject unites to each dimensional information of insertion vector
Meter, to be predicted based on transaction risk of the statistical result to the target subject.
In the exemplary embodiment, when target subject is bank's card number, the main body variable information of target subject includes bank
Card number information, bank's card number information include attaching information, that is, bank card card BIN information of target subject, can be based on bank card
Number complete information, that is, entire card number to insertion vector each dimensional information count, can also be based on the card of bank card
BIN counts each dimensional information of insertion vector.When target subject is ID card information, the main body of target subject becomes
Amount includes identification card number information, ground of the identification card number information comprising 6 expressions before attaching information, that is, identification card number of target subject
Location information can be united with each dimensional information of the complete information, that is, entire identification card number of identity-based card number to insertion vector
Meter can also count each dimensional information of insertion vector with first 6 of identity-based card number.
It for example, in the exemplary embodiment, can be based on each of target subject when target subject is bank's card number
Maximum value, minimum value or the average value of dimensional information are counted, for example, can count within a predetermined period of time with each card number
Maximum value, minimum value or the average value of corresponding transaction amount can also count corresponding with various card BIN within a predetermined period of time
Transaction amount maximum value, minimum value or average value.Further, it is also possible to statistical calculation is combined to each dimensional information,
Trading activity type dimension, transaction amount dimension, trading activity ownership dimension etc. can be combined statistical calculation, example
Such as, the total of corresponding with each card number consumer within a predetermined period of time or insurance class type of transaction transaction amount can be counted
Volume, average value, or each card BIN of statistics is corresponding consumer or insures the total value of the transaction amount of class type of transaction, is averaged
Value.
It should be noted that in this exemplary embodiment, the main body variable of target subject such as bank card and identity card is believed
Breath is preceding 6 information, but the example embodiment of the application is without being limited thereto, for example, the main body variable information of bank card can be with
It is 5 before bank's card number, the main body variable information of identity card can also be for 9 before identification card number i.e. comprising the position of year of birth
Number, as long as the main body variable information of target subject can distinguish the ownership of target subject, the application is to this without spy
It is different to limit.
Further, in the exemplary embodiment, the statistical result of each dimensional information based on insertion vector is to the target
The transaction risk of main body is predicted, for example, determining the transaction row after getting the current trading activity data of target subject
For the transaction amount under each trading activity type dimension in data maximum value whether be more than the target subject the transaction
The maximum value of historical statistics result under behavior type dimension, if being more than maximum value, it is determined that the transaction row of the target subject
To issue the user with risk prompting message with risk.According to the technical solution in the implementation of this example, even if new main body is for example new
The transaction data of card is seldom, as long as example card BIN the corresponding dimension of the new main body occurred accordingly, it will be able to determine that this is new
The transaction risk of main body improves the accuracy of the risk assessment of the trading activity of new main body.
In addition, in the exemplary embodiment, being also based on the trading activity data pair of the target subject in preset time period
Risk classifications, that is, the high risk or low-risk of the target subject are marked, and the risk classifications of the target subject based on label will be with
The corresponding insertion vector of target subject is divided into positive sample and negative sample, is carried out based on positive sample and negative sample to risk evaluation model
Training, is predicted based on transaction risk of the training result of the risk evaluation model to the current trading activity of target subject,
The risk evaluation model can be neural network model, decision-tree model or supporting vector machine model, the application to this not into
Row particular determination.
The exchange information processing method in example embodiment according to fig. 2, on the one hand, the trading activity based on target subject
Sequence determines the trading activity feature comprising attaching information of target subject, convenient for the attaching information based on target subject to target
The trading activity feature of main body is counted;On the other hand, the side that the dimensional information based on trading activity feature is embedded in by word
Formula generates insertion vector corresponding with trading activity feature, and the transaction of target subject can not only be sufficiently refined in each dimension
Behavioural characteristic, additionally it is possible to reduce calculation amount, improve data-handling efficiency;In another aspect, based on main body variable information to insertion to
Each dimensional information of amount is counted, and can be predicted based on transaction risk of the statistical result to fresh target main body, thus
It can be improved the accuracy of the risk assessment to the unconspicuous trading activity of off-note.
Fig. 3 shows the flow diagram of the generation insertion vector provided according to some embodiments of the present application.
Referring to shown in Fig. 3, in step s310, each characteristic item of the trading activity feature of target subject is clustered
Processing, the dimensional information of trading activity feature is determined based on cluster result.
It in the exemplary embodiment, can be in the case where the dimensional information of the trading activity feature of uncertain target subject
Word segmentation processing is carried out to the trading activity data in the behavior sequence of target subject, from the transaction row in the behavior sequence after participle
Related word is identified with transaction risk to extract in data, is generated corresponding term vector, is calculated the distance between each term vector,
Determine that each and transaction risk in trading activity data identifies the class cluster of related word based on the distance of calculating, by each class cluster
Dimensional information as trading activity feature.For example, cluster after class cluster may include main body variable class, transaction amount class,
Type of transaction class, exchange hour class etc., using these types as the dimensional information of trading activity feature.
It should be noted that the distance between each term vector can be Hamming distances, Euclidean distance, COS distance, but
It is that distance in the exemplary embodiment of the application is without being limited thereto, such as distance can also be mahalanobis distance, manhatton distance etc..
In step s 320, the dimensional information based on trading activity feature is reflected trading activity feature by word incorporation model
It penetrates as corresponding insertion vector.
In the exemplary embodiment, word incorporation model is LSTM model, which is a kind of deep learning model, can be with
Including input layer, hidden layer and output layer.Input layer is used to receive the trading activity feature and trading activity feature extracted
Dimensional information, hidden layer extracts corresponding dimension for the dimensional information based on trading activity feature from trading activity feature
Feature, output layer presets full articulamentum for being output to each dimensional characteristics of the trading activity feature of extraction, pre- by this
If full articulamentum generates insertion vector corresponding with trading activity feature.
In addition, in the exemplary embodiment, the historical trading behavioral data based on target subject instructs word incorporation model
Practice, based on the word incorporation model after training by the corresponding insertion vector of trading activity Feature Mapping.For example, when available default
Between multiple target subjects in section historical trading behavioral data, extracted from the historical trading behavioral data of multiple target subjects
Corresponding trading activity feature, trading activity feature and corresponding dimensional information based on target subject to word incorporation model into
Row training.
It should be noted that in this exemplary embodiment, the mode of word insertion can be with LSTM model, or
Word2Vec model or Glove model, can also for other neural network models appropriate such as Skip-Gram model or
The models such as CBOW (Continuous Bag-of-Words, continuous bag of words), this is equally within the scope of protection of this application.
Fig. 4 shows the flow diagram of the exchange information processing method provided according to other embodiments of the application.
Referring to shown in Fig. 4, in step S410, the trading activity data based on target subject generate the behavior of target subject
Sequence.
In the exemplary embodiment, which can be bank card, and the trading activity data of target subject can be receipts
Money behavioral data, the behavior sequence of target subject are gathering sequence, that is, gathering event 1 to gathering event t.
In the step s 420, the trading activity feature of the target subject is extracted from the behavior sequence of target subject.
In the exemplary embodiment, the trading activity feature of the target subject can for target subject main body variable information,
The features such as transaction amount, beneficiary information or paying party information, loco, exchange hour and type of transaction, also can wrap
Including also may include the characteristic informations such as other trading activity features appropriate such as means of payment, coupon use information.In Fig. 4
In, trading activity feature variable 1_1 to n_1, the variable 1_2 to variable n_2 of the target subject ..., variable 1_t to variable
N_t indicates that each variable indicates a feature.
In step S430, the dimensional information of the trading activity feature based on target subject is raw in such a way that word is embedded in
At insertion vector corresponding with the trading activity feature.
In the exemplary embodiment, the dimensional information of the trading activity feature of target subject may include: trading activity ownership
One of dimensional information, trading activity type dimension information, transaction amount dimensional information and exchange hour dimensional information or
It is a variety of.For example, trading activity ownership dimensional information may include bank card number information, IP address information, cell-phone number information, body
The ownership dimensional information such as part card information.
Further, word insertion is that a kind of distribution of word indicates, can be by high dimension in such a way that word is embedded in
Word space reflection is into the vector space of a low-dimensional number.In the exemplary embodiment, the dimension of trading activity feature is being determined
After information, it can be incited somebody to action by word incorporation model such as LSTM (Long Short-Term Memory, shot and long term memory) model
Insertion vector of the trading activity Feature Mapping of multidimensional to low-dimensional number corresponding with the dimensional information.Such as, it is determined that transaction row
The dimensional information being characterized includes the information of four dimensions, i.e. trading activity ownership dimensional information, trading activity type dimension letter
Breath, transaction amount dimensional information and exchange hour dimensional information, can be by word incorporation model by trading activity Feature Mapping
To the insertion vector of the four dimensions.
The LSTM model is a kind of deep learning model, may include input layer, hidden layer and output layer.Input layer is used
In the dimensional information for receiving the trading activity feature and trading activity feature extracted, hidden layer is used to be based on trading activity feature
Dimensional information extract corresponding dimensional characteristics from trading activity feature, the trading activity feature that output layer is used to extract
Each dimensional characteristics, which are output to, presets full articulamentum, presets full articulamentum by this and generates insertion corresponding with trading activity feature
Vector.
It should be noted that in this exemplary embodiment, the mode of word insertion can be with LSTM model, or
Word2Vec model or Glove model, can also for other neural network models appropriate such as Skip-Gram model or
The models such as CBOW (Continuous Bag-of-Words, continuous bag of words), this is equally within the scope of protection of this application.
In step S440, the main body variable information based on target subject unites to each dimensional information of insertion vector
Meter.
In the exemplary embodiment, when target subject is bank's card number, the main body variable information of target subject includes bank
Card number information, bank's card number information include attaching information, that is, bank card card BIN information of target subject, can be based on bank card
Number complete information, that is, entire card number to insertion vector each dimensional information count, can also be based on the card of bank card
BIN counts each dimensional information of insertion vector.When target subject is ID card information, the main body of target subject becomes
Amount includes identification card number information, ground of the identification card number information comprising 6 expressions before attaching information, that is, identification card number of target subject
Location information can be united with each dimensional information of the complete information, that is, entire identification card number of identity-based card number to insertion vector
Meter can also count each dimensional information of insertion vector with first 6 of identity-based card number.
It for example, in the exemplary embodiment, can be based on each of target subject when target subject is bank's card number
Maximum value, minimum value or the average value of dimensional information are counted, for example, can count within a predetermined period of time with each card number
Maximum value, minimum value or the average value of corresponding transaction amount can also count corresponding with various card BIN within a predetermined period of time
Transaction amount maximum value, minimum value or average value.Further, it is also possible to statistical calculation is combined to each dimensional information,
Trading activity type dimension, transaction amount dimension, trading activity ownership dimension etc. can be combined statistical calculation, example
Such as, the total of corresponding with each card number consumer within a predetermined period of time or insurance class type of transaction transaction amount can be counted
Volume, average value, or each card BIN of statistics is corresponding consumer or insures the total value of the transaction amount of class type of transaction, is averaged
Value.
Fig. 5 shows the flow diagram of the exchange information processing method provided according to the other embodiment of the application.
Referring to Figure 5, in step S510, the trading activity data of target subject within a preset period of time, example are obtained
Trading activity data such as to target subject in 3 months are deposited and are cleaned, and the noise number in trading activity data is removed
According to, and the trading activity data obtained to storage store the trading activity number of target subject for example in the form of off-line data table
According to.
In step S520, dimension behavioural characteristic corresponding with target subject, dimension row are determined based on trading activity data
Be characterized DBF (Dimension Behavior Feature, dimension behavioural characteristic) refer to the detailed data of trading activity or
The behavioural characteristic that behavior sequence is integrated in certain dimension after refining by processing.For example, available transaction
The trading activity data of target subject are mapped to corresponding dimension behavior based on the dimensional information by the dimensional information of behavioral data
Feature.
In the exemplary embodiment, step S520 may include step S524 to step S528, carry out below to these steps
Detailed description.
In step S524, the trading activity data based on target subject generate the behavior sequence of target subject.Target master
The behavior sequence of body is according to the trading activity data of the target subject in some time section of transaction time of origin sequence, example
Such as, if the trading activity data based on target subject in predetermined amount of time such as 3 months generate the behavior sequence of the target subject
Column, then can divide the trading activity data in the predetermined amount of time based on predetermined space such as 1 day, when obtaining several
Between trading activity data in section, and the trading activity data in each time interval are ranked up according to time of origin,
Generate the behavior sequence in the predetermined amount of time corresponding with the target subject.
In step S526, insertion vector corresponding with the behavior sequence of target subject is generated in such a way that word is embedded in.
Word insertion is that a kind of distribution of word indicates, can be low to one by the word space reflection of a high dimension in such a way that word is embedded in
In the vector space of dimension.In the exemplary embodiment, after the dimensional information that trading activity feature has been determined, word can be passed through
Incorporation model such as LSTM model is by the trading activity Feature Mapping of multidimensional to the insertion of low-dimensional number corresponding with the dimensional information
Vector.Such as, it is determined that the dimensional information of trading activity feature includes the information of four dimensions, i.e. trading activity ownership dimension letter
Breath, trading activity type dimension information, transaction amount dimensional information and exchange hour dimensional information, can be embedded in mould by word
Type is by the insertion vector of trading activity Feature Mapping to the four dimensions.
It is corresponding with target subject based on the corresponding insertion vector determination of behavior sequence of target subject in step S528
Dimension behavioural characteristic.In the exemplary embodiment, the main body variable information based on target subject believes each dimension of insertion vector
Breath is counted, and determines dimension behavioural characteristic corresponding with target subject based on statistical result., it is bank's card number in target subject
When, the main body variable information of target subject includes bank's card number information, and bank's card number information includes the attaching information of target subject
That is the card BIN information of bank card, can the complete information based on bank's card number, that is, entire card number to insertion vector each dimension
Information is counted, can also the card BIN based on bank card to insertion vector each dimensional information count.In target master
When body is ID card information, the main body variable of target subject includes identification card number information, and identification card number information includes target subject
Attaching information, that is, identification card number before 6 expression address informations, can with identity-based demonstrate,prove number complete information, that is, entire identity
Card number counts each dimensional information of insertion vector, can also be with first 6 of identity-based card number to each of insertion vector
A dimensional information is counted.
Next, being back to step S530, in step S530, the dimension behavioural characteristic based on target subject establishes risk
Assessment models.For example, can with the dimension behavioural characteristic of combining target main body and normal risk variable to risk evaluation model into
Row training, risk evaluation model can be the models such as neural network model, decision-tree model or supporting vector machine model, conventional
Risk variable may include the variables such as the standard deviation of transaction amount, the average value of transaction amount and abnormal behavior variable.
Further, the dimension behavioural characteristic for being also based on target subject carries out a point group to target subject, for example, in mesh
When mark main body is bank card, a point group can be carried out to target subject by the card BIN information of target subject, if the card of target subject
BIN information is identical, these target subjects are divided into a group.It, can be by new main body by carrying out a point group to target subject
Such as neocaine is divided into group appropriate, so as to improve new main body trading activity risk assessment accuracy.
In addition, the normal risk that the trading activity data based on target subject generate target subject becomes in step S540
Amount, for example, the trading activity data to target subject count, determines trade gold of the target subject in every type of transaction
The normal risks variable such as average value, standard deviation, transaction probability of volume.
Next, determining model strategy corresponding with risk evaluation model in step 550.For different risk classes
Type needs to construct different air control models and for example instead cheats model, counter usurps model, anti-cheating model etc..In every kind of risk class
Under type for different types of service use different models, by it is counter cheat model for, can be further subdivided into and take advantage of offline
Swindleness person's model transfers accounts online to account model, transfers accounts online to card mold type etc..For a type of service, model strategy can be with
Certain regular parallel form is added using multiple model parallel forms, that is, each Model score threshold value, such as: model strategy 1:
Model A score value is greater than 0.6 and transaction amount is greater than 1000 yuan;Model strategy 2 is that Model B score value is greater than 0.7 and trade gold
Volume is greater than 10000 yuan.It is assessed by using transaction risk of multiple models to target subject, risk can be further increased
The accuracy of assessment.
In step S560, based on transaction risk of the model strategy in step S550 to target subject carry out assessment and it is defeated
The result of decision out.For example, integration exports synthesis point after the output score value of different models is standardized
Value, is assessed based on transaction risk of the comprehensive scores to target subject, determines the grade of transaction risk, exports final decision
Such as indicating risk or safety instruction.
Further, the decision in the face of risk log of target subject can also be precipitated, model exports score value, risk class label etc.
Risk data further adjusts the mould of the risk evaluation model of the target subject with the risk data of the target subject based on deposition
Type strategy improves the accuracy of risk assessment.
In the example embodiment of the application, a kind of commerce information processor is additionally provided.Referring to shown in Fig. 6, the friendship
Easy information processing unit 600 include: sequence generating unit 610, feature extraction unit 620, insertion vector generation unit 630 and
Statistic unit 640.Wherein, sequence generating unit 610 generates the target master for the trading activity data based on target subject
The behavior sequence of body;Feature extraction unit 620 from the behavior sequence of the target subject for extracting the target subject
Trading activity feature, wherein the trading activity feature includes the main body variable information of the trading activity data, the main body
Variable information includes the information for indicating the attaching information of the target subject;Vector generation unit 630 is embedded in be used for based on described
The dimensional information of trading activity feature generates insertion vector corresponding with the trading activity feature in such a way that word is embedded in;
Statistic unit 640 is used to count based on each dimensional information of the main body variable information to the insertion vector, with base
It is predicted in transaction risk of the statistical result to the target subject.
In some embodiments of the present application, aforementioned schemes are based on, referring to shown in Fig. 7, the insertion vector generation unit
630 include: dimensional information determination unit 710, carries out clustering processing, base for each characteristic item to the trading activity feature
The dimensional information of the trading activity feature is determined in cluster result;Map unit 720, for being based on the trading activity feature
Dimensional information by word incorporation model by the trading activity Feature Mapping be corresponding insertion vector.
In some embodiments of the present application, aforementioned schemes are based on, referring to shown in Fig. 8, institute's predicate incorporation model is length
Phase remembers LSTM model, and the map unit 720 includes: input unit 722, for by the trading activity feature and described
The dimensional information of trading activity feature is input to the LSTM model;Feature extraction unit 724, for being based on the trading activity
The dimensional information of feature extracts corresponding dimensional characteristics from the trading activity feature by the hidden layer of the LSTM model;
Feature output unit 726 presets full articulamentum for being output to each dimensional characteristics for the trading activity feature extracted,
The corresponding insertion vector of full articulamentum generation is preset by described.
In some embodiments of the present application, it is based on aforementioned schemes, commerce information processor 600 further include: history number
According to acquiring unit, for obtaining the historical trading behavioral data of multiple target subjects;Behavioural characteristic extraction unit is used for from described
The trading activity feature of the multiple target subject is extracted in historical trading behavioral data;Training unit, for based on described more
The dimensional information of the trading activity feature of a target subject and the trading activity feature is trained the LSTM model.
In some embodiments of the present application, aforementioned schemes are based on, the statistic unit 640 includes: the first statistic unit,
It is counted for each dimensional information of the complete information based on the main body variable information to the insertion vector;And/or
Second statistic unit, the attaching information of the target subject for including based on the main body variable information to it is described be embedded in
Each dimensional information of amount is counted.
In some embodiments of the present application, aforementioned schemes are based on, the main body variable information is bank account information, the
Two statistic units are configured as: the bank identification number for including based on bank account information believes each dimension of the insertion vector
Breath is counted.
In some embodiments of the present application, aforementioned schemes are based on, the statistic unit 640 is configured as: to described embedding
Maximum value, minimum value or the average value of each dimensional information of incoming vector are counted.
In some embodiments of the present application, aforementioned schemes are based on, the statistic unit 640 includes: division unit, is used for
The target subject is divided into New Account or old debts family based on the main body variable information;First predicting unit, if for described
Target subject is New Account, then based on the corresponding system of the attaching information of the target subject that includes with the main body variable information
Meter result predicts the transaction risk of the target subject;Second predicting unit, if being old debts for the target subject
Family, then based on statistical result corresponding with the complete information of the main body variable information to the transaction risk of the target subject into
Row prediction.
In some embodiments of the present application, aforementioned schemes are based on, the statistic unit 640 is configured as: acquisition and institute
State the statistical result of each dimension of the corresponding insertion vector of main body variable;Pass through decision tree mould based on the statistical result
Type predicts the transaction risk of the target subject.
According to the commerce information processor in the example embodiment of Fig. 6, on the one hand, the trading activity based on target subject
Sequence determines the trading activity feature comprising attaching information of target subject, convenient for the attaching information based on target subject to target
The trading activity feature of main body is counted;On the other hand, the side that the dimensional information based on trading activity feature is embedded in by word
Formula generates insertion vector corresponding with trading activity feature, and the transaction of target subject can not only be sufficiently refined in each dimension
Behavioural characteristic, additionally it is possible to reduce calculation amount, improve data-handling efficiency;In another aspect, based on main body variable information to insertion to
Each dimensional information of amount is counted, and can be predicted based on transaction risk of the statistical result to fresh target main body, thus
It can be improved the accuracy of the risk assessment to the unconspicuous trading activity of off-note.
Commerce information processor provided by the embodiments of the present application can be realized each process in preceding method embodiment,
And reach identical function and effect, it is not repeated herein.
Further, the embodiment of the present application also provides a kind of Transaction Information processing equipments, as shown in Figure 9.
Transaction Information processing equipment can generate bigger difference because configuration or performance are different, may include one or one
A above processor 901 and memory 902 can store one or more storage application programs in memory 902
Or data.Wherein, memory 902 can be of short duration storage or persistent storage.The application program for being stored in memory 902 can wrap
One or more modules (diagram is not shown) are included, each module may include to a series of in Transaction Information processing equipment
Computer executable instructions.Further, processor 901 can be set to communicate with memory 902, in Transaction Information processing
The series of computation machine executable instruction in memory 902 is executed in equipment.Transaction Information processing equipment can also include one
Or more than one power supply 903, one or more wired or wireless network interfaces 904, one or more input and output
Interface 905, one or more keyboards 906 etc..
In a specific embodiment, Transaction Information processing equipment include memory and one or more
Program, perhaps more than one program is stored in memory and one or more than one program may include for one of them
One or more modules, and each module may include executable to the series of computation machine in Transaction Information processing equipment
Instruction, and be configured to execute this or more than one program by one or more than one processor to include for carrying out
Following computer executable instructions: the trading activity data based on target subject generate the behavior sequence of the target subject;From
The trading activity feature of the target subject is extracted in the behavior sequence of the target subject, wherein the trading activity feature
Main body variable information comprising the trading activity data, the main body variable information include the ownership for indicating the target subject
The information of information;Based on the dimensional information of the trading activity feature, generated and the trading activity in such a way that word is embedded in
The corresponding insertion vector of feature;It is counted based on each dimensional information of the main body variable information to the insertion vector,
To be predicted based on transaction risk of the statistical result to the target subject.
Optionally, when executed, the dimensional information based on the trading activity feature passes through computer executable instructions
The mode of word insertion generates insertion vector corresponding with the trading activity feature, comprising: to each of the trading activity feature
A characteristic item carries out clustering processing, and the dimensional information of the trading activity feature is determined based on cluster result;Based on the transaction
The trading activity Feature Mapping is corresponding insertion vector by word incorporation model by the dimensional information of behavioural characteristic.
Optionally, when executed, institute's predicate incorporation model is that shot and long term remembers LSTM model to computer executable instructions,
The trading activity Feature Mapping is corresponding embedding by word incorporation model by the dimensional information based on the trading activity feature
Incoming vector, comprising: the dimensional information of the trading activity feature and the trading activity feature is input to the LSTM mould
Type;Dimensional information based on the trading activity feature is by the hidden layer of the LSTM model from the trading activity feature
Extract corresponding dimensional characteristics;Each dimensional characteristics of the trading activity feature of extraction are output to and preset full articulamentum,
The corresponding insertion vector of full articulamentum generation is preset by described.
Optionally, computer executable instructions are when executed, further includes: obtain the historical trading row of multiple target subjects
For data;The trading activity feature of the multiple target subject is extracted from the historical trading behavioral data;Based on described more
The dimensional information of the trading activity feature of a target subject and the trading activity feature is trained the LSTM model.
Optionally, computer executable instructions when executed, are based on the main body variable information to the insertion vector
Each dimensional information counted, comprising: the complete information based on the main body variable information to it is described insertion vector it is each
A dimensional information is counted;And/or the attaching information for the target subject for based on the main body variable information including is to institute
The each dimensional information for stating insertion vector is counted.
Optionally, when executed, the main body variable information is bank account information to computer executable instructions, is based on
The attaching information for the target subject that the main body variable information includes carries out each dimensional information of the insertion vector
Statistics, comprising: the bank identification number for including based on bank account information unites to each dimensional information of the insertion vector
Meter.
Optionally, computer executable instructions when executed, unite to each dimensional information of the insertion vector
Meter, comprising: maximum value, minimum value or the average value of each dimensional information of the insertion vector are counted.
Optionally, computer executable instructions when executed, based on statistical result to the transaction wind of the target subject
It is predicted danger, comprising: the target subject is divided by New Account or old debts family based on the main body variable information;If the mesh
Mark main body is New Account, then based on the corresponding statistics of the attaching information of the target subject that includes with the main body variable information
As a result the transaction risk of the target subject is predicted;If the target subject is old debts family, it is based on and the main body
The corresponding statistical result of the complete information of variable information predicts the transaction risk of the target subject.
Optionally, computer executable instructions when executed, based on statistical result to the transaction wind of the target subject
It is predicted danger, comprising: obtain the statistical result of each dimension of the insertion vector corresponding with the main body variable;It is based on
The statistical result is predicted by transaction risk of the decision-tree model to the target subject.
Transaction Information processing equipment provided by the embodiments of the present application can be realized each process in preceding method embodiment,
And reach identical function and effect, it is not repeated herein.
In addition, the embodiment of the present application also provides a kind of storage medium, for storing computer executable instructions, a kind of tool
In the embodiment of body, which can be USB flash disk, CD, hard disk etc., the computer executable instructions of storage medium storage
When being executed by processor, be able to achieve following below scheme: the trading activity data based on target subject generate the target subject
Behavior sequence;The trading activity feature of the target subject is extracted from the behavior sequence of the target subject, wherein the friendship
Easy behavioural characteristic includes the main body variable information of the trading activity data, and the main body variable information includes to indicate the target
The information of the attaching information of main body;Based on the dimensional information of the trading activity feature, generated in such a way that word is embedded in and institute
State the corresponding insertion vector of trading activity feature;Based on the main body variable information to each dimensional information of the insertion vector
It is counted, to be predicted based on transaction risk of the statistical result to the target subject.
Optionally, the computer executable instructions of storage medium storage are based on the transaction when being executed by processor
The dimensional information of behavioural characteristic generates insertion vector corresponding with the trading activity feature in such a way that word is embedded in, comprising:
Clustering processing is carried out to each characteristic item of the trading activity feature, the trading activity feature is determined based on cluster result
Dimensional information;The trading activity Feature Mapping is by the dimensional information based on the trading activity feature by word incorporation model
Corresponding insertion vector.
Optionally, for the computer executable instructions of storage medium storage when being executed by processor, institute's predicate is embedded in mould
Type is that shot and long term remembers LSTM model, and the dimensional information based on the trading activity feature passes through word incorporation model for the transaction
Behavioural characteristic is mapped as corresponding insertion vector, comprising: by the trading activity feature and the dimension of the trading activity feature
Information input is spent to the LSTM model;Dimensional information based on the trading activity feature is hidden by the LSTM model
Layer extracts corresponding dimensional characteristics from the trading activity feature;Each dimension of the trading activity feature of extraction is special
Sign, which is output to, presets full articulamentum, presets the corresponding insertion vector of full articulamentum generation by described.
Optionally, the computer executable instructions of storage medium storage are when being executed by processor, further includes: obtain more
The historical trading behavioral data of a target subject;The friendship of the multiple target subject is extracted from the historical trading behavioral data
Easy behavioural characteristic;The dimensional information pair of trading activity feature and the trading activity feature based on the multiple target subject
The LSTM model is trained.
Optionally, the computer executable instructions of storage medium storage are based on the main body when being executed by processor
Variable information counts each dimensional information of the insertion vector, comprising: based on the complete of the main body variable information
Information counts each dimensional information of the insertion vector;And/or described in based on the main body variable information including
The attaching information of target subject counts each dimensional information of the insertion vector.
Optionally, the computer executable instructions of storage medium storage are when being executed by processor, the main body variable
Information is bank account information, and the attaching information for the target subject for including based on the main body variable information is to the insertion
Each dimensional information of vector is counted, comprising: the bank identification number for including based on bank account information to it is described be embedded in
Each dimensional information of amount is counted.
Optionally, the storage medium storage computer executable instructions when being executed by processor, to it is described be embedded in
Each dimensional information of amount is counted, comprising: to the maximum value of each dimensional information of the insertion vector, minimum value or is put down
Mean value is counted.
Optionally, the computer executable instructions of storage medium storage are based on statistical result when being executed by processor
The transaction risk of the target subject is predicted, comprising: be divided into the target subject based on the main body variable information
New Account or old debts family;If the target subject is New Account, based on the target for including with the main body variable information
The corresponding statistical result of the attaching information of main body predicts the transaction risk of the target subject;If the target subject is
Old debts family, then based on statistical result corresponding with the complete information of the main body variable information to the transaction wind of the target subject
It is predicted danger.
Optionally, the computer executable instructions of storage medium storage are based on statistical result when being executed by processor
The transaction risk of the target subject is predicted, comprising: obtain the insertion vector corresponding with the main body variable
The statistical result of each dimension;It is carried out based on the statistical result by transaction risk of the decision-tree model to the target subject
Prediction.
Computer readable storage medium provided by the embodiments of the present application can be realized each mistake in preceding method embodiment
Journey, and reach identical function and effect, it is not repeated herein.
In the 1990s, the improvement of a technology can be distinguished clearly be on hardware improvement (for example,
Improvement to circuit structures such as diode, transistor, switches) or software on improvement (improvement for method flow).So
And with the development of technology, the improvement of current many method flows can be considered as directly improving for hardware circuit.
Designer nearly all obtains corresponding hardware circuit by the way that improved method flow to be programmed into hardware circuit.Cause
This, it cannot be said that the improvement of a method flow cannot be realized with hardware entities module.For example, programmable logic device
(Programmable Logic Device, PLD) (such as field programmable gate array (Field Programmable Gate
Array, FPGA)) it is exactly such a integrated circuit, logic function determines device programming by user.By designer
Voluntarily programming comes a digital display circuit " integrated " on a piece of PLD, designs and makes without asking chip maker
Dedicated IC chip.Moreover, nowadays, substitution manually makes IC chip, this programming is also used instead mostly " is patrolled
Volume compiler (logic compiler) " software realizes that software compiler used is similar when it writes with program development,
And the source code before compiling also write by handy specific programming language, this is referred to as hardware description language
(Hardware Description Language, HDL), and HDL is also not only a kind of, but there are many kind, such as ABEL
(Advanced Boolean Expression Language)、AHDL(Altera Hardware Description
Language)、Confluence、CUPL(Cornell University Programming Language)、HDCal、JHDL
(Java Hardware Description Language)、Lava、Lola、MyHDL、PALASM、RHDL(Ruby
Hardware Description Language) etc., VHDL (Very-High-Speed is most generally used at present
Integrated Circuit Hardware Description Language) and Verilog.Those skilled in the art also answer
This understands, it is only necessary to method flow slightly programming in logic and is programmed into integrated circuit with above-mentioned several hardware description languages,
The hardware circuit for realizing the logical method process can be readily available.
Controller can be implemented in any suitable manner, for example, controller can take such as microprocessor or processing
The computer for the computer readable program code (such as software or firmware) that device and storage can be executed by (micro-) processor can
Read medium, logic gate, switch, specific integrated circuit (Application Specific Integrated Circuit,
ASIC), the form of programmable logic controller (PLC) and insertion microcontroller, the example of controller includes but is not limited to following microcontroller
Device: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320 are deposited
Memory controller is also implemented as a part of the control logic of memory.It is also known in the art that in addition to
Pure computer readable program code mode is realized other than controller, can be made completely by the way that method and step is carried out programming in logic
Controller is obtained to come in fact in the form of logic gate, switch, specific integrated circuit, programmable logic controller (PLC) and insertion microcontroller etc.
Existing identical function.Therefore this controller is considered a kind of hardware component, and to including for realizing various in it
The device of function can also be considered as the structure in hardware component.Or even, it can will be regarded for realizing the device of various functions
For either the software module of implementation method can be the structure in hardware component again.
System, device, module or the unit that above-described embodiment illustrates can specifically realize by computer chip or entity,
Or it is realized by the product with certain function.It is a kind of typically to realize that equipment is computer.Specifically, computer for example may be used
Think personal computer, laptop computer, cellular phone, camera phone, smart phone, personal digital assistant, media play
It is any in device, navigation equipment, electronic mail equipment, game console, tablet computer, wearable device or these equipment
The combination of equipment.
For convenience of description, it is divided into various units when description apparatus above with function to describe respectively.Certainly, implementing this
The function of each unit can be realized in the same or multiple software and or hardware when application.
It should be understood by those skilled in the art that, embodiments herein can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the application
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the application, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The application is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present application
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
In a typical configuration, calculating equipment includes one or more processors (CPU), input/output interface, net
Network interface and memory.
Memory may include the non-volatile memory in computer-readable medium, random access memory (RAM) and/or
The forms such as Nonvolatile memory, such as read-only memory (ROM) or flash memory (flash RAM).Memory is computer-readable medium
Example.
Computer-readable medium includes permanent and non-permanent, removable and non-removable media can be by any method
Or technology come realize information store.Information can be computer readable instructions, data structure, the module of program or other data.
The example of the storage medium of computer includes, but are not limited to phase change memory (PRAM), static random access memory (SRAM), moves
State random access memory (DRAM), other kinds of random access memory (RAM), read-only memory (ROM), electric erasable
Programmable read only memory (EEPROM), flash memory or other memory techniques, read-only disc read only memory (CD-ROM) (CD-ROM),
Digital versatile disc (DVD) or other optical storage, magnetic cassettes, tape magnetic disk storage or other magnetic storage devices
Or any other non-transmission medium, can be used for storage can be accessed by a computing device information.As defined in this article, it calculates
Machine readable medium does not include temporary computer readable media (transitory media), such as the data-signal and carrier wave of modulation.
It should also be noted that, the terms "include", "comprise" or its any other variant are intended to nonexcludability
It include so that the process, method, commodity or the equipment that include a series of elements not only include those elements, but also to wrap
Include other elements that are not explicitly listed, or further include for this process, method, commodity or equipment intrinsic want
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including described want
There is also other identical elements in the process, method of element, commodity or equipment.
It will be understood by those skilled in the art that embodiments herein can provide as method, system or computer program product.
Therefore, complete hardware embodiment, complete software embodiment or embodiment combining software and hardware aspects can be used in the application
Form.It is deposited moreover, the application can be used to can be used in the computer that one or more wherein includes computer usable program code
The shape for the computer program product implemented on storage media (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.)
Formula.
The application can describe in the general context of computer-executable instructions executed by a computer, such as program
Module.Generally, program module includes routines performing specific tasks or implementing specific abstract data types, programs, objects, group
Part, data structure etc..The application can also be practiced in a distributed computing environment, in these distributed computing environments, by
Task is executed by the connected remote processing devices of communication network.In a distributed computing environment, program module can be with
In the local and remote computer storage media including storage equipment.
All the embodiments in this specification are described in a progressive manner, same and similar portion between each embodiment
Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.Especially for system reality
For applying example, since it is substantially similar to the method embodiment, so being described relatively simple, related place is referring to embodiment of the method
Part explanation.
The above description is only an example of the present application, is not intended to limit this application.For those skilled in the art
For, various changes and changes are possible in this application.All any modifications made within the spirit and principles of the present application are equal
Replacement, improvement etc., should be included within the scope of the claims of this application.