CN109784733A - User credit prediction technique, device, electronic equipment and storage medium - Google Patents
User credit prediction technique, device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN109784733A CN109784733A CN201910049615.6A CN201910049615A CN109784733A CN 109784733 A CN109784733 A CN 109784733A CN 201910049615 A CN201910049615 A CN 201910049615A CN 109784733 A CN109784733 A CN 109784733A
- Authority
- CN
- China
- Prior art keywords
- credit
- prediction
- initial
- user
- characteristic data
- 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
Abstract
Present disclose provides a kind of user credit prediction technique, device, electronic equipment and computer readable storage mediums, belong to field of artificial intelligence.This method comprises: obtaining the initial characteristic data of target user, the initial characteristic data includes at least one of face characteristic data, expressive features data and motion characteristic data;The initial characteristic data of the target user is converted into target feature vector;The credit prediction model for being utilized respectively multiple credit sections handles the target feature vector, obtains the target user in the initial prediction in the multiple credit section;Based on each initial prediction, the final predicted value of the target user is determined.The disclosure needs not rely upon external credit data, and the quick predict of user credit may be implemented, improve efficiency, and prediction result accuracy with higher.
Description
Technical field
This disclosure relates to field of artificial intelligence, in particular to a kind of user credit prediction technique, user credit prediction
Device, electronic equipment and computer readable storage medium.
Background technique
With the development of computer technology, internet also has been more and more widely used in financial field.Wherein, it participates in
Internet financial consumption, financing lease, the service of receiving credit card or the user of other financial services are more and more, for ensuring funds
Melt the interests of service enterprise, reduces financial risks, it is necessary to take effective user credit evaluation method.
Existing user credit evaluation method usually requires user and submits a series of external datas, such as people's row reference number
According to, main strategies data, assets prove etc., pass through external data evaluate user credit standing.As it can be seen that this method is for outer
Portion's data have stronger dependence, and generally also need to audit external data, to increase needed for credit appraisal
Time, reduce efficiency.
It should be noted that information is only used for reinforcing the reason to the background of the disclosure disclosed in above-mentioned background technology part
Solution, therefore may include the information not constituted to the prior art known to persons of ordinary skill in the art.
Summary of the invention
Present disclose provides a kind of user credit prediction technique, user credit prediction meanss, electronic equipment and computers can
Storage medium is read, and then overcomes the problems, such as that existing user credit evaluation method required time is too long at least to a certain extent.
Other characteristics and advantages of the disclosure will be apparent from by the following detailed description, or partially by the disclosure
Practice and acquistion.
According to one aspect of the disclosure, a kind of user credit prediction technique is provided, comprising: obtain the original of target user
Characteristic, the initial characteristic data include face characteristic data, in expressive features data and motion characteristic data at least
It is a kind of;The initial characteristic data of the target user is converted into target feature vector;It is utilized respectively the letter in multiple credit sections
The target feature vector is handled with prediction model, obtains the target user in the initial of the multiple credit section
Predicted value;Based on each initial prediction, the final predicted value of the target user is determined.
In a kind of exemplary embodiment of the disclosure, the method also includes: obtain the original spy of multiple historical users
Levy data and credit appraisal value;The initial characteristic data of the historical user is converted into sampling feature vectors;It is gone through according to described
The historical user is divided into the multiple credit section by the credit appraisal value of history user;For each credit section,
Using the sampling feature vectors and credit appraisal value training machine learning model of historical user in the credit section, obtain described
The credit prediction model in credit section.
In a kind of exemplary embodiment of the disclosure, the method also includes: the credit for obtaining the target user is commented
Value;If the credit appraisal value of the target user and final predicted value are not in same credit section, by the mesh
Mark user is added in the historical user, and the credit in credit section belonging to target user described in re -training predicts mould
Type.
It is described to be based on each initial prediction in a kind of exemplary embodiment of the disclosure, determine that the target is used
The final predicted value at family includes: the departure degree based on each initial prediction in the corresponding credit section, is determined
The final predicted value of the target user.
It is described to be based on each initial prediction in corresponding credit section in a kind of exemplary embodiment of the disclosure
Interior departure degree, determine the target user final predicted value include: calculate separately each initial prediction with it is described
The departure degree of the median in the corresponding credit section of initial prediction;The smallest initial prediction of the departure degree is determined
For the final predicted value.
It is described to be based on each initial prediction in corresponding credit section in a kind of exemplary embodiment of the disclosure
Interior departure degree, determine the target user final predicted value include: calculate separately each initial prediction with it is described
The departure degree of the median in the corresponding credit section of initial prediction;The initial prediction is determined according to the departure degree
Weight;Each initial prediction is weighted, the final predicted value is obtained.
In a kind of exemplary embodiment of the disclosure, the initial characteristic data includes the primitive character number of multiple indexs
According to;It includes: respectively by each index that the initial characteristic data by the target user, which is converted to target feature vector,
Initial characteristic data is transformed into the range of [0,1];Using each index as a dimension, the target for generating various dimensions is special
Levy vector.
According to one aspect of the disclosure, a kind of user credit prediction meanss are provided, comprising: data acquisition module is used for
The initial characteristic data of target user is obtained, the initial characteristic data includes face characteristic data, expressive features data and moves
Make at least one of characteristic;Vector conversion module, for the initial characteristic data of the target user to be converted to mesh
Mark feature vector;Initial predicted module, for being utilized respectively the credit prediction model in multiple credit sections to the target signature
Vector is handled, and obtains the target user in the initial prediction in the multiple credit section;Final prediction module, is used for
Based on each initial prediction, the final predicted value of the target user is determined.
In a kind of exemplary embodiment of the disclosure, the data acquisition module is also used to obtain multiple historical users'
Initial characteristic data and credit appraisal value;The vector conversion module is also used to turn the initial characteristic data of the historical user
It is changed to sampling feature vectors;Described device further include: model training module, for the credit appraisal value according to the historical user
The historical user is divided into the multiple credit section, and for each credit section, utilizes the credit area
The sampling feature vectors and credit appraisal value training machine learning model of interior historical user, obtain the credit in the credit section
Prediction model.
In a kind of exemplary embodiment of the disclosure, the data acquisition module is also used to obtain the target user's
Credit appraisal value;If the model training module is also used to the credit appraisal value of the target user and final predicted value is not located
In in same credit section, then the target user is added in the historical user, and target user described in re -training
The credit prediction model in affiliated credit section.
In a kind of exemplary embodiment of the disclosure, the final prediction module is also used to based on each initial predicted
It is worth the departure degree in the corresponding credit section, determines the final predicted value of the target user.
In a kind of exemplary embodiment of the disclosure, the final prediction module is also used to calculate separately each described initial
The departure degree of the median in predicted value credit corresponding with initial prediction section, and the departure degree is the smallest
Initial prediction is determined as the final predicted value.
In a kind of exemplary embodiment of the disclosure, the final prediction module is also used to calculate separately each described initial
The departure degree of the median in predicted value credit corresponding with initial prediction section, determines institute according to the departure degree
The weight of initial prediction is stated, and each initial prediction is weighted, obtains the final predicted value.
In a kind of exemplary embodiment of the disclosure, the initial characteristic data includes the primitive character number of multiple indexs
According to;In the range of the vector conversion module is also used to that the initial characteristic data of each index is transformed into [0,1] respectively, and
Using each index as a dimension, the target feature vector of various dimensions is generated.
According to one aspect of the disclosure, a kind of electronic equipment is provided, comprising: processor;And memory, for storing
The executable instruction of the processor;Wherein, the processor is configured to above-mentioned to execute via the executable instruction is executed
Method described in any one.
According to one aspect of the disclosure, a kind of computer readable storage medium is provided, computer program is stored thereon with,
The computer program realizes method described in above-mentioned any one when being executed by processor.
The exemplary embodiment of the disclosure has the advantages that
Its target feature vector is generated according to the initial characteristic data of target user, then inputs multiple credit sections respectively
Credit prediction model obtains multiple initial predictions about target user's credit, finally according to initial prediction corresponding
Departure degree in credit section determines the final predicted value of target user's credit.On the one hand, it proposes a kind of according to user's
The method that face feature, expression or action data carry out credit prediction needs not rely upon external credit data, and carries out in user
Credit prediction can be realized after face action certification, shorten the process of user credit evaluation, improve efficiency.On the other hand,
Multiple credit sections are set, and the model for being utilized respectively each credit section carries out initial predicted, and is based on initial predicted as a result, obtaining
To the final prediction result of optimization, to refine the process of credit prediction, the accuracy of prediction result is improved.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The disclosure can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the disclosure
Example, and together with specification for explaining the principles of this disclosure.It should be evident that the accompanying drawings in the following description is only the disclosure
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 shows a kind of flow chart of user credit prediction technique of the present exemplary embodiment;
Fig. 2 shows the sub-process figures of user credit prediction technique a kind of in the present exemplary embodiment;
Fig. 3 shows the flow chart of another user credit prediction technique in the present exemplary embodiment;
Fig. 4 shows a kind of structural block diagram of user credit prediction meanss in the present exemplary embodiment;
Fig. 5 shows a kind of electronic equipment for realizing the above method in the present exemplary embodiment;
Fig. 6 shows a kind of computer readable storage medium for realizing the above method in the present exemplary embodiment.
Specific embodiment
Example embodiment is described more fully with reference to the drawings.However, example embodiment can be with a variety of shapes
Formula is implemented, and is not understood as limited to example set forth herein;On the contrary, thesing embodiments are provided so that the disclosure will more
Fully and completely, and by the design of example embodiment comprehensively it is communicated to those skilled in the art.Described feature, knot
Structure or characteristic can be incorporated in any suitable manner in one or more embodiments.
The exemplary embodiment of the disclosure provides firstly a kind of user credit prediction technique, can be applied to following scene
In: terminal user is in register account number (such as open a bank account, stock is opened an account) or carries out sensitive operation (such as payment, Shen
Please provide a loan, change cell-phone number binding etc.) when, it needs to complete specified face action according to the guidance of program, server can acquire
User data in action process carries out credit prediction to it according to data.Therefore, the executing subject of the present exemplary embodiment can
To be server.Refering to what is shown in Fig. 1, the user credit prediction technique may comprise steps of S110~S140:
Step S110 obtains the initial characteristic data of target user.
Wherein, target user is the user for needing to carry out credit prediction;Initial characteristic data can be face characteristic data,
Any one or more in expressive features data and motion characteristic data is carrying out face action certification for characterizing target user
When characteristic information, specifically, to can be the characterization facial appearance such as shape of face, eyes distance, eye sizes special for face characteristic data
The data of sign, expressive features data can be mouth shapes and change the characterization facial expression spies such as numerical value, eye size change numerical value
The data of sign, motion characteristic data can be characterization head or the face actions such as head shaking direction, shake time, shake amplitude
The data of feature;It in practical applications, can be by expressive features number since human face expression also can be considered that the fine motion of face is made
It is handled according to a kind of data are merged into motion characteristic data.
In the present embodiment, the initial characteristic data of target user can be acquired by the photographing module of terminal, and be sent to clothes
Business device.For example, user makes according to the guidance of program in verification process and controls the required movements such as shake the head, nod, smile
Or expression, terminal collect the initial characteristic data during this, send it to server.
The initial characteristic data of target user is converted to target feature vector by step S120.
Server can carry out certain pretreatment after obtaining initial characteristic data to it, such as screen out nothing therein
With data or can not letter data etc., certain statistics processing can also be carried out to it, is then converted into the form of vector, with
Convenient for subsequent processing.Usual initial characteristic data includes multiple data, can be arranged, then be converted in particular order
For target feature vector.The numerical value of each dimension can indicate that target user is a certain in verification process in target feature vector
The motion characteristic of aspect, therefore target feature vector contains whole motion characteristic information.
It should be noted that if in initial characteristic data including nonumeric type data, it is above-mentioned pre- in vector conversion
Treatment process include numeralization processing is carried out to nonumeric type data, such as: the initial characteristic data of shape of face is usually shape of face
Approximate shapes preset each shape mapped numerical value, and numeralization processing may be implemented, and if shape of face is ellipse, then convert
For numerical value [1], circle be [2], it is rectangular be [3], triangle is [4], up-side down triangle is [5], diamond shape is [6], rectangle is [7]
Etc..In addition, in pretreatment, company larger for the numeric type data in initial characteristic data, especially numeric distribution range
The continuous stronger data of property can also carry out sliding-model control to it by dividing numberical range, initial characteristic data is converted
For the vector value in target feature vector.A large amount of initial characteristic data can be usually enumerated in advance, carry out range division, then
Each range is provided to the mapping relations between vector value, with the initial characteristic data of eyes distance for example, can be set 3
A numberical range: 39%~41%, 41%~43%, 43%~45%, the primitive character between 39%~41% range
Data are converted into vector value [0], are converted into vector value [2] between 41%~43% range, are located at 43%~45% model
Vector value [4] etc. is converted between enclosing.
Step S130, the credit prediction model for being utilized respectively multiple credit sections handle target feature vector, obtain
Initial prediction to target user in multiple credit sections.
In the present embodiment, credit prediction and evaluation, such as setting 0~100 can be carried out to user by way of numerical value
Credit value range, numerical value is bigger, and expression credit is better, on this basis, entire credit value range can be divided into more
A credit section, such as in 0~100 numberical range, 0~25 can be delimited as the first credit section, 25~50 be second
Credit section, 50~75 be third credit section, and 75~100 be the 4th credit section etc..Certainly, the present embodiment is for credit
The quantity in section and the numberical range in each credit section are not specially limited.
Each credit section has corresponding credit prediction model, usually completes trained machine learning model.
The input of credit prediction model is to characterize the feature vector of user action data, is exported as the credit number in respective credit section
Value.Target feature vector is inputted respectively in each credit prediction model, each credit prediction model exports an initial predicted
Value.Initial prediction is to carry out the obtained credit value of tentative prediction, the letter in each credit section to the credit of target user
It should be in the credit section with the initial prediction that prediction model exports.
In one exemplary embodiment, it can have the portion of overlapping between credit section adjacent in multiple credit sections
Point, such as the case where above-mentioned credit value range is 0~100, it can be set 0~40 the first credit section, the of 30~70
Two credit sections, 60~100 third credit section, then with 30~40 between the first credit section and the second credit section
Lap has 60~70 lap between the second credit section and third credit section.The letter of reduplicative forms is set
It is advantageously reduced with section and occurs the case where Sparse at credit interval border, increase the robustness of credit prediction technique,
Improve the accuracy of result.
Step S140 is based on each initial prediction, determines the final predicted value of target user.
Wherein, final predicted value is the final result of target user's credit prediction.It can be determined in each initial prediction
Optimal one, as final predicted value, such as the prediction of each initial prediction that can be obtained according to each credit prediction model
Probability determines the matching degree of each initial prediction with corresponding credit section, and the highest initial prediction of matching degree is determined as
Final predicted value can also carry out average or average weighted calculating, using result as final predicted value to each initial prediction
Etc., the present embodiment is not specially limited this.
Based on above description, in the present example embodiment, its target is generated according to the initial characteristic data of target user
Feature vector, then the credit prediction model in multiple credit sections is inputted respectively, it obtains multiple about the initial of target user's credit
Predicted value finally determines the final pre- of target user's credit according to departure degree of the initial prediction in corresponding credit section
Measured value.On the one hand, a kind of method that credit prediction is carried out according to the face feature of user, expression or action data, nothing are proposed
External credit data need to be depended on, and credit prediction can be realized after user carries out face action certification, shortens user's letter
With the process of evaluation, efficiency is improved.On the other hand, multiple credit sections are set, the model in each credit section is utilized respectively
Initial predicted is carried out, and based on initial predicted as a result, the final prediction result optimized, to refine the mistake of credit prediction
Journey improves the accuracy of prediction result.
In one exemplary embodiment, refering to what is shown in Fig. 2, user credit prediction technique can also include for obtaining credit
Step S210~S240 of prediction model:
Step S210 obtains the initial characteristic data and credit appraisal value of multiple historical users;
The initial characteristic data of historical user is converted to sampling feature vectors by step S220;
Historical user is divided into multiple credit sections according to the credit appraisal value of historical user by step S230;
Step S240 utilizes the sampling feature vectors and letter of historical user in the credit section for each credit section
With evaluation of estimate training machine learning model, the credit prediction model in the credit section is obtained.
Wherein, credit appraisal value is the real credit behavior according to user or the obtained credit number of external credit data
Value, be different from above-mentioned credit initial prediction and final predicted value, can be considered the actual value of user credit, generate the time one
As be later than user carry out face action certification time, that is, be later than credit prediction time.In the present embodiment, credit is obtained and has commented
The user of value can be considered historical user, and historical user generally also carried out face action and recognize before obtaining credit appraisal value
Card, therefore initial characteristic data was also generated, using the initial characteristic data of historical user-credit appraisal value as sample data
Group can train credit prediction model.By acquiring the initial characteristic data and credit appraisal value of a large amount of historical users, can obtain
Obtain a large amount of sample data group.
Before training pattern, need to carry out certain pretreatment, the i.e. process of step S220 and S230.The original of historical user
Beginning characteristic needs to be converted to sampling feature vectors, and the concrete mode of conversion can be consistent with step S120, guarantees sample
The form consistency of feature vector and target feature vector.Historical user needs to classify according to credit section, makes each letter
It is served only for training the credit prediction model in the credit section with the data of the historical user in section, credit prediction model can be improved
Specific aim and accuracy.
In the training process of step S240, sampling feature vectors are input in credit prediction model, pass through iteration tune
The parameter of integral mould becomes closer to the numerical value of output in credit appraisal value, finally reaches certain accuracy rate in verifying.
Each credit prediction model is mutually independent model, and training process can be mutually indepedent, the letter that a training can also be completed
Use prediction model as the initial model of another credit prediction model, it, can also be in each letter to solve the problems, such as the cold start-up of model
With branch's parameter etc. is shared between prediction model, the present embodiment is not specially limited this.
In one exemplary embodiment, above-mentioned machine learning model may include supporting vector machine model, logistic regression mould
Type or neural network model.These three types of models are adapted to using vector as input, using successional numerical value as output, excavation
Non-linear relation between each feature.The credit prediction model in each credit section can be same class model, be also possible to difference
Class model.
In one exemplary embodiment, user credit prediction technique can with the following steps are included:
Obtain the credit appraisal value of target user;
If the credit appraisal value of target user and final predicted value are not in same credit section, by target user
It is added in historical user, and the credit prediction model in credit section belonging to training objective user again.
According to foregoing teachings, for target user after carrying out face action certification, server can carry out credit prediction to it,
Prediction result, that is, above-mentioned final predicted value;Hereafter during target user receives to service, if producing real credit row
For or have submitted external credit data etc., server can carry out actual credit appraisal to it again, or due to target user
Special event (such as in violation of rules and regulations, breaking one's promise) occurs, credit appraisal, evaluation result, that is, credit appraisal value manually are made to target user.
Therefore there may be certain difference between credit appraisal value and final predicted value, if difference is excessive, so that two values are not
In same credit section, illustrate prediction result inaccuracy, it can be with re -training credit prediction model.
In re -training, target user can be added in above-mentioned historical user, by the primitive character of target user
Data-credit appraisal value is added to original sample data and concentrates as new sample data group.It should be noted that target
Credit section belonging to user refers to the credit section where the credit appraisal value of target user, rather than where its final predicted value
Credit section, when re -training, the new sample data group that target user generates is only used for training the credit area belonging to it
Between credit prediction model, and to other credit prediction models without re -training.Mould can be corrected or be optimized to re -training
Parameter in type further increases its accuracy rate.
By the re -training of credit prediction model, credit prediction model guide service strategy may be implemented --- service plan
Slightly generate service data --- service data feeds back to the closed loop mechanism of credit prediction model, allow credit prediction model not
It is disconnected to update and perfect, the accuracy of raising credit prediction.
In one exemplary embodiment, it can have the portion of overlapping between credit section adjacent in multiple credit sections
Point.Correspondingly, a historical user may be divided into two even more credit sections in step S230, increase
The reusability of sample data, and in step S140, it is easier to characterize the corresponding credit section of each initial prediction
Departure degree, so that entire credit prediction algorithm has better robustness.In addition, the credit appraisal value of target user may belong to
It, then, can be right if when final predicted value is not in any one in these credit sections in two even more than credit section
The credit prediction model in these credit sections all carries out re -training.
Fig. 3 shows a kind of entire flow of user credit prediction technique in the present exemplary embodiment.Refering to what is shown in Fig. 3,
The initial characteristic data for obtaining target user first, can be face feature data, expressive features data, motion characteristic data
Deng being converted into target feature vector, and input multiple credit prediction models respectively;Each credit prediction model output is initial pre-
Measured value, then optimal value is determined from initial prediction, as final predicted value;Can using final predicted value as prediction result,
Final predicted value can also be mapped on corresponding credit section, which credit section prediction target user is belonged into as pre-
Survey result.Wherein, credit prediction model is obtained based on historical use data training, in training, first by historical user
It is divided into each credit section, then in each credit section, historical use data is converted into sampling feature vectors-credit appraisal
The data group of value, to be trained to model, credit prediction model A, the credit prediction model B and credit obtained in Fig. 3 is predicted
MODEL C etc..In addition, if in the follow-up process, the credit area of actual credit appraisal value and prediction that target user obtains
Between it is different, then the data of target user can be used as to training data, re -training credit prediction model, to carry out more accurately
Credit prediction.To realize the whole process of user credit prediction and prediction algorithm update.
In one exemplary embodiment, step S140 can be realized by following steps:
Departure degree based on each initial prediction in corresponding credit section determines the final prediction of target user
Value.
In the present embodiment, the real credit numerical value of target user be should be in some or multiple credit sections, can be with
It is called target credit section, then passes through the credit prediction model processing target feature in the credit section other than target credit section
When vector, the initial prediction of output and the matching degree in the credit section are poor, are usually expressed as initial prediction and deviate the letter
At the upper limit or lower limit with the median in section, or positioned at the credit section.It therefore can be according to each initial prediction right
Departure degree in the credit section answered calculates the matching degree of target user and each credit section, and then determines target user's credit
Final predicted value.Specific processing method can be illustrated with him by following example:
In one exemplary embodiment, the departure degree based on each initial prediction in corresponding credit section determines
The final predicted value of target user, can with specifically includes the following steps:
Calculate separately the departure degree of each initial prediction with the median in corresponding credit section;
It will deviate from the smallest initial prediction of degree and be determined as final predicted value.
I.e. for n credit section Q1~Qn, it is utilized respectively its corresponding credit prediction model model (1)~model
(n) target feature vector A is handled, obtains initial predicted value set P={ p1, p2 ... pn };It, can for any pi ∈ P
To calculate its departure degree:
Bi=| pi-med (Qi) |;
Wherein, bi indicates that the departure degree of i-th initial prediction and its credit section, med (Qi) are credit section Qi
Median, the usually average value of the limit value of Qi and lower numerical limit.The inclined of each initial prediction is calculated in the method
From degree b1~bn, if wherein bj is minimum, its corresponding initial prediction pj is determined as final predicted value F.In brief,
It is considered that the smallest initial prediction of departure degree is optimal value, as final predicted value.
In one exemplary embodiment, the departure degree based on each initial prediction in corresponding credit section determines
The final predicted value of target user, can also with specifically includes the following steps:
Calculate separately the departure degree of each initial prediction with the median in corresponding credit section;
The weight of initial prediction is determined according to departure degree;
Each initial prediction is weighted, final predicted value is obtained.
Specifically, the departure degree b1 of the median in the corresponding credit section each initial prediction p1~pn is calculated
After~bn, a kind of illustrative method for calculating final predicted value F is as follows:
It can will deviate from the root reciprocal of degreeWeight coefficient as each initial prediction.In addition, can also
With using the weight coefficient of other concrete forms, departure degree is bigger to embody, the smaller Computing Principle of weight, this reality
It applies example and this is not specially limited.
You need to add is that some possible corresponding credit section matching degree is very low in initial prediction,
It shows as initial prediction to be at the upper limit or the lower limit in credit section, then final prediction is being calculated by average weighted method
When value, this part of initial prediction can be deleted in advance, then be weighted.
In one exemplary embodiment, it after obtaining final predicted value, can map that on corresponding credit section,
Prediction target user is belonged into the credit section as final prediction result.
Further, it after obtaining about the credit prediction result of target user, can be taken according to the result subsequent
Service strategy, such as decide whether examine loan, the amount of money that examination & approval are provided a loan, whether core sends out credit card, institute's core sends out the volume of credit card
Degree etc..
In one exemplary embodiment, initial characteristic data may include the initial characteristic data of multiple indexs;Correspondingly,
Step S120 can be realized by following steps:
The initial characteristic data of each index is transformed into the range of [0,1] respectively;
Using each index as a dimension, the target feature vector of various dimensions is generated.
Wherein, the index of initial characteristic data can be the attribute of data, such as above-mentioned shape of face, eyes distance, eyes ruler
Very little, mouth shapes change numerical value, eye size change numerical value, head shaking direction, shake time, shake amplitude etc..It will be original
Characteristic is transformed into the range of [0,1], can take normalized mode, such as calculates the initial characteristic data of each index
Calculating etc. is normalized relative to the ratio of maximum, or using initial characteristic data of the normalized function to each index.
After the numerical value being transformed into [0,1] range, the data of each index can be converted in target feature vector
The numerical value of specific one dimension.Index corresponding to the number of dimensions and each dimension of target feature vector can be set, such as
Fruit initial characteristic data does not include the data of some index, then the numerical value of corresponding dimension can be set as 0.
The above process is further illustrated: assuming that obtaining original spy of multiple target users in verification process
Data are levied, quantity, movement range, reaction time etc. including carrying out required movement, initial characteristic data can be recorded and be located
It manages as follows:
$ Data=array (
Array (2=> 0.43,3=> 0.12,1284=> 0.2 ...),
Array (1=> 0.22,5=> 0.02,394=> 0.11 ...),
);
The data of each index are carried out with the conversion of numberical range, and is arranged in a certain order, it then can will be every
The initial characteristic data of a target user is converted to target feature vector:
A1=[0.43,0.12,0.2 ...];
A2=[0.22,0.02,0.11 ...].
By way of the above-mentioned normalized by initial characteristic data, the numeric distribution of each index can be balanced, favorably
Unified processing is carried out in subsequent, operand is reduced, further increases efficiency.
The exemplary embodiment of the disclosure additionally provides a kind of user credit prediction meanss, refering to what is shown in Fig. 4, the device
400 may include: data acquisition module 410, and for obtaining the initial characteristic data of target user, initial characteristic data includes people
At least one of face characteristic, expressive features data and motion characteristic data;Vector conversion module 420 is used for target
The initial characteristic data of user is converted to target feature vector;Initial predicted module 430, for being utilized respectively multiple credit sections
Credit prediction model target feature vector is handled, obtain target user in the initial prediction in multiple credit sections;
Final prediction module 440 determines target user for the departure degree based on each initial prediction in corresponding credit section
Final predicted value.
In one exemplary embodiment, data acquisition module can be also used for obtaining the primitive character number of multiple historical users
According to credit appraisal value;Vector conversion module can be also used for being converted to the initial characteristic data of historical user sample characteristics to
Amount;User credit prediction meanss can also include: model training module, for according to the credit appraisal value of historical user by history
User is divided into multiple credit sections, and for each credit section, utilizes the sample characteristics of historical user in credit section
Vector and credit appraisal value training machine learning model, obtain the credit prediction model in credit section.
In one exemplary embodiment, data acquisition module can be also used for obtaining the credit appraisal value of target user;Mould
If type training module can be also used for the credit appraisal value of target user and final predicted value is not in same credit section,
Then target user is added in historical user, and the credit prediction model in credit section belonging to training objective user again.
In one exemplary embodiment, final prediction module can be also used for based on each initial prediction in corresponding credit
Departure degree in section determines the final predicted value of target user.
In one exemplary embodiment, final prediction module can be also used for calculating separately each initial prediction and initial pre-
The departure degree of the median in the corresponding credit section of measured value, and will deviate from the smallest initial prediction of degree and be determined as finally in advance
Measured value.
In one exemplary embodiment, final prediction module can be also used for calculating separately each initial prediction and initial pre-
The departure degree of the median in the corresponding credit section of measured value determines the weight of initial prediction and right according to departure degree
Each initial prediction is weighted, and obtains final predicted value.
In one exemplary embodiment, initial characteristic data may include the initial characteristic data of multiple indexs;Vector turns
Mold changing block can be also used in the range of the initial characteristic data of each index is transformed into [0,1] respectively, and with each index work
For a dimension, the target feature vector of various dimensions is generated.
The detail of each module in above-mentioned apparatus has carried out in corresponding method section Example detailed
Illustrate, therefore repeats no more.
The exemplary embodiment of the disclosure additionally provides a kind of electronic equipment that can be realized the above method.
Person of ordinary skill in the field it is understood that various aspects of the disclosure can be implemented as system, method or
Program product.Therefore, various aspects of the disclosure can be with specific implementation is as follows, it may be assumed that complete hardware embodiment, complete
The embodiment combined in terms of full Software Implementation (including firmware, microcode etc.) or hardware and software, can unite here
Referred to as circuit, " module " or " system ".
The electronic equipment 500 of this exemplary embodiment according to the disclosure is described referring to Fig. 5.What Fig. 5 was shown
Electronic equipment 500 is only an example, should not function to the embodiment of the present disclosure and use scope bring any restrictions.
As shown in figure 5, electronic equipment 500 is showed in the form of universal computing device.The component of electronic equipment 500 can wrap
It includes but is not limited to: at least one above-mentioned processing unit 510, at least one above-mentioned storage unit 520, the different system components of connection
The bus 530 of (including storage unit 520 and processing unit 510), display unit 540.
Wherein, storage unit is stored with program code, and program code can be executed with unit 510 processed, so that processing is single
Member 510 executes the step described in above-mentioned " illustrative methods " part of this specification according to the various illustrative embodiments of the disclosure
Suddenly.For example, processing unit 510 can execute step S110~S140 shown in FIG. 1, step shown in Fig. 2 can also be executed
S210~S240 etc..
Storage unit 520 may include the readable medium of volatile memory cell form, such as Random Access Storage Unit
(RAM) 521 and/or cache memory unit 522, it can further include read-only memory unit (ROM) 523.
Storage unit 520 can also include program/utility 524 with one group of (at least one) program module 525,
Such program module 525 includes but is not limited to: operating system, one or more application program, other program modules and
It may include the realization of network environment in program data, each of these examples or certain combination.
Bus 530 can be to indicate one of a few class bus structures or a variety of, including storage unit bus or storage
Cell controller, peripheral bus, graphics acceleration port, processing unit use any bus structures in a variety of bus structures
Local bus.
Electronic equipment 500 can also be with one or more external equipments 700 (such as keyboard, sensing equipment, bluetooth equipment
Deng) communication, can also be enabled a user to one or more equipment interact with the electronic equipment 500 communicate, and/or with make
Any equipment (such as the router, modulation /demodulation that the electronic equipment 500 can be communicated with one or more of the other calculating equipment
Device etc.) communication.This communication can be carried out by input/output (I/O) interface 550.Also, electronic equipment 500 can be with
By network adapter 560 and one or more network (such as local area network (LAN), wide area network (WAN) and/or public network,
Such as internet) communication.As shown, network adapter 560 is communicated by bus 530 with other modules of electronic equipment 500.
It should be understood that although not shown in the drawings, other hardware and/or software module can not used in conjunction with electronic equipment 500, including but not
Be limited to: microcode, device driver, redundant processing unit, external disk drive array, RAID system, tape drive and
Data backup storage system etc..
Through the above description of the embodiments, those skilled in the art is it can be readily appreciated that example described herein is implemented
Mode can also be realized by software realization in such a way that software is in conjunction with necessary hardware.Therefore, according to the disclosure
The technical solution of embodiment can be embodied in the form of software products, which can store non-volatile at one
Property storage medium (can be CD-ROM, USB flash disk, mobile hard disk etc.) in or network on, including some instructions are so that a calculating
Equipment (can be personal computer, server, terminal installation or network equipment etc.) is executed according to the exemplary implementation of the disclosure
The method of example.
The exemplary embodiment of the disclosure additionally provides a kind of computer readable storage medium, and being stored thereon with can be realized
The program product of this specification above method.In some possible embodiments, various aspects of the disclosure can also be realized
For a kind of form of program product comprising program code, when program product is run on the terminal device, program code is used for
Execute terminal device described in above-mentioned " illustrative methods " part of this specification according to the various exemplary embodiment party of the disclosure
The step of formula.
It is produced refering to what is shown in Fig. 6, describing the program according to the exemplary embodiment of the disclosure for realizing the above method
Product 600, can be using portable compact disc read only memory (CD-ROM) and including program code, and can set in terminal
It is standby, such as run on PC.However, the program product of the disclosure is without being limited thereto, in this document, readable storage medium storing program for executing can
With to be any include or the tangible medium of storage program, the program can be commanded execution system, device or device use or
Person is in connection.
Program product can be using any combination of one or more readable mediums.Readable medium can be readable signal Jie
Matter or readable storage medium storing program for executing.Readable storage medium storing program for executing for example can be but be not limited to electricity, magnetic, optical, electromagnetic, infrared ray or partly lead
System, device or the device of body, or any above combination.More specific example (the non exhaustive column of readable storage medium storing program for executing
Table) it include: the electrical connection with one or more conducting wires, portable disc, hard disk, random access memory (RAM), read-only storage
Device (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read only memory (CD-
ROM), light storage device, magnetic memory device or above-mentioned any appropriate combination.
Computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
In carry readable program code.The data-signal of this propagation can take various forms, including but not limited to electromagnetic signal,
Optical signal or above-mentioned any appropriate combination.Readable signal medium can also be any readable Jie other than readable storage medium storing program for executing
Matter, the readable medium can send, propagate or transmit for by instruction execution system, device or device use or and its
The program of combined use.
The program code for including on readable medium can transmit with any suitable medium, including but not limited to wirelessly, have
Line, optical cable, RF etc. or above-mentioned any appropriate combination.
Can with any combination of one or more programming languages come write for execute the disclosure operation program
Code, programming language include object oriented program language-Java, C++ etc., further include conventional process
Formula programming language-such as " C " language or similar programming language.Program code can be calculated fully in user
It executes in equipment, partly execute on a user device, executing, as an independent software package partially in user calculating equipment
Upper part executes on a remote computing or executes in remote computing device or server completely.It is being related to remotely counting
In the situation for calculating equipment, remote computing device can pass through the network of any kind, including local area network (LAN) or wide area network
(WAN), it is connected to user calculating equipment, or, it may be connected to external computing device (such as utilize ISP
To be connected by internet).
In addition, above-mentioned attached drawing is only the schematic theory of the processing according to included by the method for disclosure exemplary embodiment
It is bright, rather than limit purpose.It can be readily appreciated that the time that above-mentioned processing shown in the drawings did not indicated or limited these processing is suitable
Sequence.In addition, be also easy to understand, these processing, which can be, for example either synchronously or asynchronously to be executed in multiple modules.
It should be noted that although being referred to several modules or list for acting the equipment executed in the above detailed description
Member, but this division is not enforceable.In fact, according to an exemplary embodiment of the present disclosure, above-described two or
More multimode or the feature and function of unit can embody in a module or unit.Conversely, above-described one
A module or the feature and function of unit can be to be embodied by multiple modules or unit with further division.
Those skilled in the art after considering the specification and implementing the invention disclosed here, will readily occur to its of the disclosure
His embodiment.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or
Adaptive change follow the general principles of this disclosure and including the undocumented common knowledge in the art of the disclosure or
Conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the disclosure are by claim
It points out.
It should be understood that the present disclosure is not limited to the precise structures that have been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.The scope of the present disclosure is only limited by the attached claims.
Claims (10)
1. a kind of user credit prediction technique characterized by comprising
The initial characteristic data of target user is obtained, the initial characteristic data includes face characteristic data, expressive features data
At least one of with motion characteristic data;
The initial characteristic data of the target user is converted into target feature vector;
The credit prediction model for being utilized respectively multiple credit sections handles the target feature vector, obtains the target
Initial prediction of the user in the multiple credit section;
Based on each initial prediction, the final predicted value of the target user is determined.
2. the method according to claim 1, wherein the method also includes:
Obtain the initial characteristic data and credit appraisal value of multiple historical users;
The initial characteristic data of the historical user is converted into sampling feature vectors;
The historical user is divided into the multiple credit section according to the credit appraisal value of the historical user;
For each credit section, the sampling feature vectors and credit appraisal value of historical user in the credit section are utilized
Training machine learning model obtains the credit prediction model in the credit section.
3. according to the method described in claim 2, it is characterized in that, the method also includes:
Obtain the credit appraisal value of the target user;
If the credit appraisal value of the target user and final predicted value are not in same credit section, by the target
User is added in the historical user, and the credit prediction model in credit section belonging to target user described in re -training.
4. determining the mesh the method according to claim 1, wherein described be based on each initial prediction
Mark user final predicted value include:
Departure degree based on each initial prediction in the corresponding credit section, determines the target user most
Whole predicted value.
5. according to the method described in claim 4, it is characterized in that, described be based on each initial prediction in corresponding credit
Departure degree in section determines that the final predicted value of the target user includes:
Calculate separately the departure degree of the median in each initial prediction credit corresponding with initial prediction section;
The smallest initial prediction of the departure degree is determined as the final predicted value.
6. according to the method described in claim 4, it is characterized in that, described be based on each initial prediction in corresponding credit
Departure degree in section determines that the final predicted value of the target user includes:
Calculate separately the departure degree of the median in each initial prediction credit corresponding with initial prediction section;
The weight of the initial prediction is determined according to the departure degree;
Each initial prediction is weighted, the final predicted value is obtained.
7. the method according to claim 1, wherein the initial characteristic data includes the original spy of multiple indexs
Levy data;
The initial characteristic data by the target user is converted to target feature vector and includes:
The initial characteristic data of each index is transformed into the range of [0,1] respectively;
Using each index as a dimension, the target feature vector of various dimensions is generated.
8. a kind of user credit prediction meanss characterized by comprising
Data acquisition module, for obtaining the initial characteristic data of target user, the initial characteristic data includes face characteristic
At least one of data, expressive features data and motion characteristic data;
Vector conversion module, for the initial characteristic data of the target user to be converted to target feature vector;
Initial predicted module, the credit prediction model for being utilized respectively multiple credit sections carry out the target feature vector
Processing, obtains the target user in the initial prediction in the multiple credit section;
Final prediction module determines the final predicted value of the target user for being based on each initial prediction.
9. a kind of electronic equipment characterized by comprising
Processor;And
Memory, for storing the executable instruction of the processor;
Wherein, the processor is configured to require 1-7 described in any item via executing the executable instruction and carry out perform claim
Method.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
Claim 1-7 described in any item methods are realized when being executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910049615.6A CN109784733A (en) | 2019-01-18 | 2019-01-18 | User credit prediction technique, device, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910049615.6A CN109784733A (en) | 2019-01-18 | 2019-01-18 | User credit prediction technique, device, electronic equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109784733A true CN109784733A (en) | 2019-05-21 |
Family
ID=66501596
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910049615.6A Pending CN109784733A (en) | 2019-01-18 | 2019-01-18 | User credit prediction technique, device, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109784733A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111191891A (en) * | 2019-12-18 | 2020-05-22 | 东方微银科技(北京)有限公司 | Evaluation accuracy calibration method and device, electronic device and storage medium |
-
2019
- 2019-01-18 CN CN201910049615.6A patent/CN109784733A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111191891A (en) * | 2019-12-18 | 2020-05-22 | 东方微银科技(北京)有限公司 | Evaluation accuracy calibration method and device, electronic device and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109919684A (en) | For generating method, electronic equipment and the computer readable storage medium of information prediction model | |
CA3118313A1 (en) | Methods and systems for improving machines and systems that automate execution of distributed ledger and other transactions in spot and forward markets for energy, compute, storage and other resources | |
CN109983491B (en) | Method and apparatus for applying artificial intelligence to money collection by using voice input | |
CN110197315A (en) | Methods of risk assessment, device and its storage medium | |
US11605081B2 (en) | Method and device applying artificial intelligence to send money by using voice input | |
US20190180255A1 (en) | Utilizing machine learning to generate recommendations for a transaction based on loyalty credits and stored-value cards | |
CN108846077A (en) | Semantic matching method, device, medium and the electronic equipment of question and answer text | |
US20210117889A1 (en) | Co-operative resource pooling system | |
CN110020862A (en) | A kind of business risk appraisal procedure, device and computer readable storage medium | |
CN110264330A (en) | Credit index calculating method, device, computer readable storage medium | |
CN109035028A (en) | Intelligence, which is thrown, cares for strategy-generating method and device, electronic equipment, storage medium | |
CN107273979A (en) | The method and system of machine learning prediction are performed based on service class | |
CN108492170A (en) | A kind of repayment method, computer readable storage medium and terminal device | |
US20210004809A1 (en) | Fraud prevention for payment instruments | |
US20190362354A1 (en) | Real-time updating of predictive analytics engine | |
CN109711696A (en) | Enterprise's methods of marking, device, medium and electronic equipment | |
CN110348471A (en) | Exception object recognition methods, device, medium and electronic equipment | |
US11663662B2 (en) | Automatic adjustment of limits based on machine learning forecasting | |
US11640627B2 (en) | Artificial intelligence based service recommendation | |
CN109784733A (en) | User credit prediction technique, device, electronic equipment and storage medium | |
US11748759B1 (en) | Using transaction data to present search results | |
KR102485874B1 (en) | AR Vision Service System to Prevent Failure Cost | |
Cao et al. | Developing actionable trading agents | |
US20240152584A1 (en) | Authentication data aggregation | |
US20230401417A1 (en) | Leveraging multiple disparate machine learning model data outputs to generate recommendations for the next best action |
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 |