CN108182634A - A kind of training method for borrowing or lending money prediction model, debt-credit Forecasting Methodology and device - Google Patents
A kind of training method for borrowing or lending money prediction model, debt-credit Forecasting Methodology and device Download PDFInfo
- Publication number
- CN108182634A CN108182634A CN201810098233.8A CN201810098233A CN108182634A CN 108182634 A CN108182634 A CN 108182634A CN 201810098233 A CN201810098233 A CN 201810098233A CN 108182634 A CN108182634 A CN 108182634A
- Authority
- CN
- China
- Prior art keywords
- debt
- credit
- model
- user
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
This application provides a kind of training method for borrowing or lending money prediction model, debt-credit prediction model and device, this method to include：The history debt-credit data of multiple users based on acquisition, determine debt-credit attributive character and lend-borrow action feature；Using the debt-credit attributive character as independent variable, using the lend-borrow action feature as dependent variable, build at least two preset models and be trained；Based on the prediction result that at least two preset model obtains, fusion treatment is carried out at least two preset model using preset model fusion method, obtains debt-credit prediction model.
Description
Technical field
This application involves data analysis technique fields, in particular to a kind of training method for borrowing or lending money prediction model, borrow
Borrow Forecasting Methodology and device.
Background technology
At present, the platforms such as major shopping, finance, internet finance are each provided with debt-credit function, so as to user in fund not
Oneself interested article can be bought when sufficient.When user is borrowed or lent money, each large platform can be according to the personal base of user
Multiple dimensions such as this information, academic information, social information, job information, history loan information carry out the repayment energy of personal loan
Force estimation it is, personal credit scores, determines whether to borrow or lend money and borrow or lend money amount according to credit scoring.In order to expand finance
Business, each large platform also can the data based on user to user's future whether carry out debt-credit predict, still, existing prediction
Model is relatively single, and the accuracy of prediction is not high, be unfavorable for each large platform commence business expand work.
Invention content
In view of this, the application be designed to provide a kind of debt-credit prediction model training method, debt-credit Forecasting Methodology and
Device, for solving the problems, such as that the debt-credit accuracy of prediction user in the prior art is relatively low.
In a first aspect, the embodiment of the present application provides a kind of training method for borrowing or lending money prediction model, this method includes：
The history debt-credit data of multiple users based on acquisition, determine debt-credit attributive character and lend-borrow action feature；
Using the debt-credit attributive character as independent variable, using the lend-borrow action feature as dependent variable, structure at least two
A preset model is simultaneously trained；
Based on the prediction result that at least two preset model obtains, using preset model fusion method to it is described at least
Two preset models carry out fusion treatment, obtain debt-credit prediction model.
Optionally, the prediction result obtained based at least two preset model, using preset model fusion method pair
At least two preset model carries out fusion treatment, specifically includes：
Using the prediction result of at least two preset model as independent variable, using lend-borrow action feature as dependent variable,
Structure Fusion Model is simultaneously trained.
Optionally, using the debt-credit attributive character as independent variable, using the lend-borrow action feature as dependent variable, structure
At least two preset models are simultaneously trained, and are specifically included：
The history debt-credit data of multiple users based on acquisition, determine that each user corresponds to the feature of debt-credit attributive character
The characteristic value of value and corresponding lend-borrow action feature；
Determine at least two preset models；And
For each preset model, each user is corresponded to value of the characteristic value as independent variable of debt-credit attributive character, it will
Value of the characteristic value of corresponding lend-borrow action feature as dependent variable, is trained the preset model, obtains completing training extremely
Few two preset models.
Optionally, at least two preset model includes：Neural network prediction model；
For each preset model, each user is corresponded to value of the characteristic value as independent variable of debt-credit attributive character, it will
Value of the characteristic value of corresponding lend-borrow action feature as dependent variable, is trained the preset model, specifically includes：
For multiple neural network models, following training operation is performed respectively, wherein, multiple neural networks have different
The neural network number of plies：
Each user is corresponded to value of the characteristic value as independent variable of debt-credit attributive character, by corresponding lend-borrow action feature
Value of the characteristic value as dependent variable, is trained Current Situation of Neural Network model, obtain Current Situation of Neural Network model parameter and
For weighing the index value of the pre-set level of model prediction accuracy；
The corresponding neural network model of the highest index value of forecasting accuracy will be characterized as finally determining neural network
Prediction model.
Optionally, the history debt-credit data of the multiple user include going through for the first historical time section of the multiple user
History borrows or lends money the history debt-credit data of the second historical time section of data and the multiple user, and the first historical time section earlier than
Second historical time section；
The history debt-credit data of multiple users based on acquisition, determine that each user corresponds to the feature of debt-credit attributive character
The characteristic value of value and corresponding lend-borrow action feature, specifically includes：
The history debt-credit data of the first historical time section of the multiple user based on acquisition, determine that each user corresponds to
Borrow or lend money the characteristic value of attributive character；
The history debt-credit data of the second historical time section of the multiple user based on acquisition, determine that each user corresponds to
The characteristic value of lend-borrow action feature.
Second aspect, the embodiment of the present application provide a kind of debt-credit Forecasting Methodology, and this method includes：
The history debt-credit data of user to be predicted based on acquisition, determine the debt-credit attributive character of the user to be predicted
Characteristic value；
Using the characteristic value of determining debt-credit attributive character as independent variable, the debt-credit prediction mould that input determines as described above
Type predicts the debt-credit probability of the user to be predicted.
The third aspect, the embodiment of the present application provide a kind of training device for borrowing or lending money prediction model, which includes：
First processing module, for multiple users based on acquisition history borrow or lend money data, determine debt-credit attributive character with
And lend-borrow action feature；
Training module, for using the debt-credit attributive character as independent variable, using the lend-borrow action feature as because of change
Amount builds at least two preset models and is trained；
Second processing module, for the prediction result obtained based at least two preset model, using preset model
Fusion method carries out fusion treatment at least two preset model, obtains debt-credit prediction model.
Fourth aspect, the embodiment of the present application provide a kind of debt-credit prediction meanss, which includes：
Processing module borrows or lends money data for the history of the user to be predicted based on acquisition, determines the user's to be predicted
Borrow or lend money the characteristic value of attributive character；
Prediction module, for the value of the characteristic value of debt-credit attributive character that will determine as independent variable, input is determined as above
Debt-credit prediction model, predict the debt-credit probability of the user to be predicted.
5th aspect, the embodiment of the present application provide a kind of computer equipment and include memory, processor and be stored in institute
The computer program that can be run on memory and on the processor is stated, the processor performs real during the computer program
The step of showing above-mentioned method.
6th aspect, the embodiment of the present application provide a kind of computer equipment and include memory, processor and be stored in institute
The computer program that can be run on memory and on the processor is stated, the processor performs real during the computer program
The step of showing method described above.
Training method, debt-credit Forecasting Methodology and the device of debt-credit prediction model provided by the embodiments of the present application, are borrowed by described in
Attributive character is borrowed as independent variable, using the lend-borrow action feature as dependent variable, at least two preset models is built and carries out
Training, based on the prediction result that at least two preset model obtains, using preset model fusion method to described at least two
A preset model carries out fusion treatment, obtains debt-credit prediction model, carries out debt-credit prediction relative to single model, improve progress
The accuracy of prediction is borrowed or lent money, is convenient for accurately finance marketing.
For the above-mentioned purpose of the application, feature and advantage is enable to be clearer and more comprehensible, preferred embodiment cited below particularly, and coordinate
Appended attached drawing, is described in detail below.
Description of the drawings
It, below will be to needed in the embodiment attached in order to illustrate more clearly of the technical solution of the embodiment of the present application
Figure is briefly described, it should be understood that the following drawings illustrates only some embodiments of the application, therefore is not construed as pair
The restriction of range, for those of ordinary skill in the art, without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 is a kind of flow diagram of training method for borrowing or lending money prediction model provided by the embodiments of the present application；
Fig. 2 is a kind of schematic diagram of the prediction accuracy of logistic regression prediction model provided by the embodiments of the present application；
The schematic diagram of Fig. 3 correspondences between a kind of neural network number of plies provided by the embodiments of the present application and index value；
Fig. 4 is a kind of schematic diagram of the prediction accuracy of neural network prediction model provided by the embodiments of the present application；
Fig. 5 is a kind of flow diagram for borrowing or lending money Forecasting Methodology provided by the embodiments of the present application；
Fig. 6 is a kind of structure diagram of training device for borrowing or lending money prediction model provided by the embodiments of the present application；
Fig. 7 is a kind of structure diagram for borrowing or lending money prediction meanss provided by the embodiments of the present application；
Fig. 8 is a kind of structure diagram of computer equipment 800 provided by the embodiments of the present application；
Fig. 9 is a kind of structure diagram of computer equipment 900 provided by the embodiments of the present application.
Specific embodiment
Purpose, technical scheme and advantage to make the embodiment of the present application are clearer, below in conjunction with the embodiment of the present application
The technical solution in the embodiment of the present application is clearly and completely described in middle attached drawing, it is clear that described embodiment is only
It is some embodiments of the present application, instead of all the embodiments.The application being usually described and illustrated herein in the accompanying drawings is real
Applying the component of example can be configured to arrange and design with a variety of different.Therefore, below to the application's for providing in the accompanying drawings
The detailed description of embodiment is not intended to limit claimed scope of the present application, but is merely representative of the selected reality of the application
Apply example.Based on embodiments herein, institute that those skilled in the art are obtained under the premise of creative work is not made
There is other embodiment, shall fall in the protection scope of this application.
As shown in Figure 1, the embodiment of the present application provides a kind of training method for borrowing or lending money prediction model, this method includes following
Step：
S101, the history debt-credit data of multiple users based on acquisition, determines that debt-credit attributive character and lend-borrow action are special
Sign；
Specifically, history debt-credit data can be obtained from the default platform such as shopping platforms such as Taobao, Jingdone district, generally comprise use
Family information, sequence information, the click information of user, borrowing balance, loaning bill relevant information etc..The specific history of each user is borrowed
It borrows data and can refer to following table：
Debt-credit attributive character and lend-borrow action feature may be generally based upon above-mentioned history debt-credit data and determine, wherein, debt-credit
Whether the purchasing power of attributive character characterization user and the borrowing balance information of user, lend-borrow action characteristic present user borrow money.
S102, using the debt-credit attributive character as independent variable, using the lend-borrow action feature as dependent variable, structure is extremely
Few two preset models are simultaneously trained；
Specifically, preset model can be Logic Regression Models, neural network prediction model etc., and the application refuses this
Limitation.
Using the debt-credit attributive character as independent variable, using the lend-borrow action feature as dependent variable, structure at least two
A preset model is simultaneously trained, and specifically includes following steps：
Step 1: the history debt-credit data of multiple users based on acquisition, determine that each user corresponds to debt-credit attributive character
Characteristic value and corresponding lend-borrow action feature characteristic value；
Step 2: determine at least two preset models；And
Step 3: for each preset model, each user is corresponded into the characteristic value of debt-credit attributive character as independent variable
Value, using the characteristic value of corresponding lend-borrow action feature as the value of dependent variable, which is trained, obtains completing to instruct
At least two experienced preset models.
Here, the history debt-credit data of acquisition are usually data of multiple users in multiple historical time sections, multiple users
History debt-credit data including the multiple user the first historical time section history debt-credit data and multiple users second
The history debt-credit data of historical time section, and the first historical time section is earlier than the second historical time section；
Data are borrowed or lent money in the history of multiple users based on acquisition, determine that each user corresponds to the feature of debt-credit attributive character
The characteristic value of value and corresponding lend-borrow action feature, specifically includes following steps：
Step 1: the history debt-credit data of the first historical time section of the multiple user based on acquisition, determine each
User corresponds to the characteristic value of debt-credit attributive character；
Step 2: the history debt-credit data of the second historical time section of the multiple user based on acquisition, determine each
User corresponds to the characteristic value of lend-borrow action feature.
Further, the debt-credit prediction model of training completes training in actual use in the embodiment of the present invention, Ke Yigen
Predicted according to whether the history loan information of user can occur lend-borrow action to the user in future time, therefore, into
Row training when, as debt-credit attributive character characteristic value can come from the first historical time section history borrow or lend money data, as by means of
The characteristic value for borrowing behavioural characteristic can come from the history debt-credit data of the second historical time section.Preferably, the first historical time section
Time interval between the second historical time section is first time interval, and the history that debt-credit prediction model uses when in use is borrowed
Period where borrowing information and needs predict whether that the time interval between the future time section of debt-credit is the second time interval, the
One time interval is consistent with the second time interval.
Specifically, the period can be continuous number of days, month or time etc., and the first historical time section generally comprises multiple
Period, e.g., continuous 3 months.Second historical time section is generally 1 period, e.g., 1 month.First historical time section and
Second historical time section can be continuous the period, still, the first historical time section earlier than the second historical time section, for example,
First historical time section can be July, August, September, and the second historical time section can be October, then, can be with when prediction
Based on September, October, November loan information, the lend-borrow action in December is predicted (assuming that July-November is historical time
Section, December is future time section).
In specific implementation, it after the history debt-credit data for obtaining the second historical time section of user, can be borrowed according to history
The history credit amount in data is borrowed, determines that each user corresponds to the characteristic value of lend-borrow action feature, if for example, user A is 10
Month borrow money 10000 yuan, at this point, the characteristic value of the corresponding lend-borrow action features of user A be 1, if user A October borrow money 0
Member, at this point, the characteristic value of the corresponding lend-borrow action features of user A is 0.
After the history debt-credit data for obtaining the first historical time section of multiple users, it can be borrowed based on the above-mentioned history of acquisition
Borrow the following derivative variable of data structure：
After above-mentioned derivative variable is obtained, can the data volume based on each derivative variable, using preset characteristic processing
Method handles the derivative variable, obtains the characteristic value of debt-credit attributive character.Wherein, preset characteristic processing method packet
Include evidence weight (Weight of evidence, WOE) branch mailbox processing method, dummy variable processing method, at decision tree branch mailbox
Reason method, etc. frequency divisions case processing method, standardization processing method etc., above-mentioned derivative variable is handled using the above method
Process has detailed introduction in the prior art, is no longer excessively illustrated herein.
Illustrated by taking initial amount as an example, for example, initial amount includes 1000,5000,10000,20000, X users couple
The initial amount answered is 5000, and after being handled using dummy variable processing method the initial amount of X user, what is obtained is above-mentioned
The corresponding characteristic value for borrowing or lending money attributive character of initial amount is (0,1,0,0).
When two preset models are respectively logistic regression prediction model and neural network prediction model, by determining debt-credit
Value of the characteristic value of attributive character as independent variable, using the characteristic value of lend-borrow action feature as the value of dependent variable, inputs respectively
Logistic regression prediction model and neural network prediction model are trained, and obtain completing logistic regression prediction model and the god of training
Through Network Prediction Model.
After the logistic regression prediction model for obtaining completing training, correspondence obtains the finger for weighing model prediction accuracy
Scale value.Wherein, which can be that Andrei Kolmogorov-Si meter Nuo Fu examines KS (Kolmogorov-Smirnov test)
Area AUC (Area under Curve of receiver operating below check value, recipient's operating characteristic curve
Characteristic curve) index value.The prediction result of logistic regression prediction model is as follows：
Wherein, weight of the corresponding values of Estimate for each independent variable in logistic regression prediction model.
The KS check values of logistic regression prediction model are that 0.587, AUC index values are 0.856, can refer to Fig. 2.
The principle of logistic regression prediction model is as follows：
Multivariate linear model is：H (x)=a0+a1x1+a2x2+…+anxn
Wherein, dependent variables of the h (x) for multivariate linear model, a0、a1、……anFor the weight of independent variable, x1、x2、……xn
Independent variable for multivariate linear model.
Classified using multivariate linear model to article, preset threshold values, it is then that all dependent variable h (x) are big
It is divided into one kind in the sample of threshold values, others are divided into another kind of.But this mode has a problem that, since the value of h (x) is to appoint
Size of anticipating, the selection of threshold values is a difficult thing, and for the ease of the selection of threshold values, it is normalized.
If threshold values is：T, then
Wherein, ha(x) to utilize multivariate linear model prediction result；
Assuming that：
a0=a0- t, aTX=a0+a1x1+…+anxn
Using S types (sigmoid) function pair, it is normalized herein.
If at this point, estimate parameter using square smallest error function, since the function after normalization is non-convex function, therefore
And its minimum value cannot be found using gradient descent method.But use the method estimation model parameter of Maximum-likelihood estimation.
Due to being two classification, it can be assumed that：
P (y=1 | xi)=ha(xi), p (y=0 | xi)=1-ha(xi)
Wherein, P (y=1 | xi) it is the probability that prediction result is 1；
ha(xi) be
So likelihood function is：
Wherein, h (xi) be
M is positive integer；
Log-likelihood function L (a)：
To L (a) maximizings, the estimated value of a is obtained.
When preset model is neural network prediction model, since neural network has the different neural network numbers of plies, needle
To multiple neural network models, following training operation is performed respectively：
Each user is corresponded to value of the characteristic value as independent variable of debt-credit attributive character, by corresponding lend-borrow action feature
Value of the characteristic value as dependent variable, is trained Current Situation of Neural Network model, obtain Current Situation of Neural Network model parameter and
For weighing the index value of the pre-set level of model prediction accuracy；
The corresponding neural network model of the highest index value of forecasting accuracy will be characterized as finally determining neural network
Prediction model.
In specific implementation, the number of plies of neural network could be provided as most 10 layers, and since one layer, often one layer of increase is right
A neural network model is answered, value of the characteristic value as independent variable of attributive character will be borrowed or lent money, by the feature of lend-borrow action feature
Be worth value as dependent variable, input above-mentioned neural network model respectively and be trained, obtain each neural network model parameter and
The index value of model prediction accuracy is weighed, e.g., which can be KS check values, AUC index values.In practical applications,
KS check values and AUC index values can be obtained simultaneously, with reference to figure 3.
The accuracy of the bigger characterization model prediction of index value is higher, therefore, by the corresponding neural network of highest index value
Model is as finally determining neural network prediction model, as shown in figure 3, when the neural network number of plies is 9, corresponding index value is most
Greatly, which is final neural network prediction model.
S103, based on the prediction result that at least two preset model obtains, using preset model fusion method to institute
It states at least two preset models and carries out fusion treatment, obtain debt-credit prediction model.
In the prediction result obtained based at least two preset model, using preset model fusion method to it is described extremely
When few two preset models carry out fusion treatment, following steps are specifically included：
Using the prediction result of at least two preset model as independent variable, using lend-borrow action feature as dependent variable,
Structure Fusion Model is simultaneously trained.
Here, fusion treatment is carried out to preset model, finally obtains debt-credit prediction model, can preferably be predicted excellent
Degree increases the accuracy of debt-credit prediction.Model Fusion has following two methods：
Method one：Model accumulates (Model stacking), votes the prediction result of each preset model, is taken using minority
From most principles, usually the prediction result of several preset models being weighted and is averaging, weights are directly proportional to model prediction goodness,
It is inversely proportional with the uncertainty of model.
Method two：Model integrated (Model ensemble), using the prediction result of each preset model as output valve, instruction
Practice a new grader, then, by the use of trained grader prediction result as borrow or lend money the final prediction result of prediction model.
In specific implementation, two preset models are respectively logistic regression prediction model and neural network prediction model, are incited somebody to action
Value of the characteristic value of the corresponding debt-credit attributive character of first historical time section as independent variable, by the debt-credit of the second historical time section
Value of the characteristic value of behavioural characteristic as dependent variable, respectively input logic regressive prediction model obtain the first prediction result, input
Finally determining neural network prediction model obtains the second prediction result.
Further, using the first prediction result and the second prediction result as the value of independent variable, by the second historical time section
Value of the lend-borrow action feature as dependent variable, the Fusion Model for inputting structure are trained, and finally obtain debt-credit prediction model, together
When can also obtain the KS check values of Knowledge Verification Model accuracy and AUC index values.
Prediction result is as follows：
Wherein, Estimate is the weighted value of the independent variable of above-mentioned debt-credit prediction model.
KS check values are：0.5903507, AUC index value is：0.856, with reference to figure 4, it is seen that predict mould using above-mentioned debt-credit
The accuracy of type prediction is higher than the accuracy individually predicted using logistic regression prediction model or neural network prediction model.
As shown in figure 5, the embodiment of the present application provides a kind of debt-credit Forecasting Methodology, specifically include：
S501, the history of the user to be predicted based on acquisition borrow or lend money data, determine the debt-credit attribute of the user to be predicted
The characteristic value of feature；
S502, using the characteristic value of determining debt-credit attributive character as independent variable, the above-mentioned determining debt-credit prediction mould of input
Type predicts the debt-credit probability of the user to be predicted.
Specifically, the history debt-credit data of the user to be predicted of acquisition can be 1 historical time section or multiple history
The history debt-credit data of period；The process of the characteristic value of the debt-credit attribute of user to be predicted is determined based on history debt-credit data,
Process above can be referred to, is no longer excessively illustrated herein.
In one embodiment, the probability borrowed or lent money to predict user to be predicted in the April, what can be obtained treats
It predicts the history debt-credit data in January of user, February, March, data is borrowed or lent money according to the above-mentioned history of acquisition, are determined to be predicted
The characteristic value of the debt-credit attributive character of user, the process are described in detail above, are not being illustrated excessively herein.
Using the characteristic value of determining debt-credit attributive character as the value of independent variable, the above-mentioned two that input training is completed respectively
Preset model obtains the corresponding prediction result of each preset model, using obtained prediction result as the value of independent variable, has inputted
Into trained debt-credit prediction model, the debt-credit probability of user to be predicted is finally obtained.
As shown in fig. 6, the embodiment of the present application provides a kind of training device for borrowing or lending money prediction model, which includes：
First processing module 61 borrows or lends money data for the history of multiple users based on acquisition, determines debt-credit attributive character
And lend-borrow action feature；
Training module 62, for using the debt-credit attributive character as independent variable, using the lend-borrow action feature as because
Variable builds at least two preset models and is trained；
Second processing module 63, for the prediction result obtained based at least two preset model, using default mould
Type fusion method carries out fusion treatment at least two preset model, obtains debt-credit prediction model.
Optionally, the Second processing module 63 is specifically used for：
Using the prediction result of at least two preset model as independent variable, using lend-borrow action feature as dependent variable,
Structure Fusion Model is simultaneously trained.
Optionally, the training module 63 specifically includes：
The history debt-credit data of multiple users based on acquisition, determine that each user corresponds to the feature of debt-credit attributive character
The characteristic value of value and corresponding lend-borrow action feature；
Determine at least two preset models；And
For each preset model, each user is corresponded to value of the characteristic value as independent variable of debt-credit attributive character, it will
Value of the characteristic value of corresponding lend-borrow action feature as dependent variable, is trained the preset model, obtains completing training extremely
Few two preset models.
Optionally, at least two preset model includes：Neural network prediction model；
The training module 62 is additionally operable to：For each preset model, each user is corresponded to the spy of debt-credit attributive character
Value of the value indicative as independent variable using the characteristic value of corresponding lend-borrow action feature as the value of dependent variable, carries out the preset model
Training, specifically includes：
For multiple neural network models, following training operation is performed respectively, wherein, multiple neural networks have different
The neural network number of plies：
Each user is corresponded to value of the characteristic value as independent variable of debt-credit attributive character, by corresponding lend-borrow action feature
Value of the characteristic value as dependent variable, is trained Current Situation of Neural Network model, obtain Current Situation of Neural Network model parameter and
For weighing the index value of the pre-set level of model prediction accuracy；
The corresponding neural network model of the highest index value of forecasting accuracy will be characterized as finally determining neural network
Prediction model.
Optionally, the history debt-credit data of the multiple user include going through for the first historical time section of the multiple user
History borrows or lends money the history debt-credit data of the second historical time section of data and the multiple user, and the first historical time section earlier than
Second historical time section；
Optionally, the first processing module 61 is specifically used for：
The history debt-credit data of the first historical time section of the multiple user based on acquisition, determine that each user corresponds to
Borrow or lend money the characteristic value of attributive character；
The history debt-credit data of the second historical time section of the multiple user based on acquisition, determine that each user corresponds to
The characteristic value of lend-borrow action feature.
As shown in fig. 7, the embodiment of the present application provides a kind of debt-credit prediction meanss, which includes：
Processing module 71 borrows or lends money data for the history of the user to be predicted based on acquisition, determines the user to be predicted
Debt-credit attributive character characteristic value；
Prediction module 72, for the value of the characteristic value of debt-credit attributive character that will determine as independent variable, input is as above-mentioned
The debt-credit prediction model that the training device of debt-credit prediction model determines predicts the debt-credit probability of the user to be predicted.
Corresponding to the training method of the debt-credit prediction model in Fig. 1, the embodiment of the present application additionally provides a kind of computer and sets
Standby 800, as shown in figure 8, the equipment includes memory 801, processor 802 and is stored on the memory 801 and can be at this
The computer program run on reason device 802, wherein, above-mentioned processor 802 realizes above-mentioned debt-credit when performing above computer program
The training method of prediction model.
Specifically, above-mentioned memory 801 and processor 802 can be general memory and processor, do not do have here
Body limits, and when the computer program of 802 run memory 801 of processor storage, is able to carry out above-mentioned debt-credit prediction model
Training method, so as to solve the problems, such as to predict that the debt-credit accuracy of user is relatively low in the prior art, by the debt-credit attributive character
As independent variable, using the lend-borrow action feature as dependent variable, build at least two preset models and be trained, based on institute
State the prediction result that at least two preset models obtain, using preset model fusion method at least two preset model into
Row fusion treatment obtains debt-credit prediction model, debt-credit prediction is carried out relative to single model, improves the standard for carrying out debt-credit prediction
Exactness is convenient for accurately finance marketing.
Corresponding to the training method of the debt-credit prediction model in Fig. 1, the embodiment of the present application additionally provides a kind of computer can
Storage medium is read, computer program is stored on the computer readable storage medium, when which is run by processor
The step of performing the training method of above-mentioned debt-credit prediction model.
Specifically, which can be general storage medium, such as mobile disk, hard disk, on the storage medium
Computer program when being run, be able to carry out the training method of above-mentioned debt-credit prediction model, it is pre- in the prior art so as to solve
The problem of debt-credit accuracy of survey user is relatively low, using the debt-credit attributive character as independent variable, by the lend-borrow action feature
As dependent variable, build at least two preset models and be trained, the prediction obtained based at least two preset model
As a result, carrying out fusion treatment at least two preset model using preset model fusion method, debt-credit prediction model is obtained,
Debt-credit prediction is carried out relative to single model, improves the accuracy for carrying out debt-credit prediction, is convenient for accurately finance marketing.
Corresponding to the debt-credit Forecasting Methodology in Fig. 5, the embodiment of the present application additionally provides a kind of computer equipment 900, such as Fig. 9
Shown, which includes memory 901, processor 902 and is stored on the memory 901 and can be transported on the processor 902
Capable computer program, wherein, above-mentioned processor 902 realizes above-mentioned debt-credit Forecasting Methodology when performing above computer program.
Specifically, above-mentioned memory 901 and processor 902 can be general memory and processor, do not do have here
Body limits, and when the computer program of 902 run memory 901 of processor storage, is able to carry out above-mentioned debt-credit Forecasting Methodology, from
And solve the problems, such as that the debt-credit accuracy of prediction user in the prior art is relatively low, the accuracy for carrying out debt-credit prediction is improved, just
It markets in carrying out accurately finance.
Corresponding to the debt-credit Forecasting Methodology in Fig. 5, the embodiment of the present application additionally provides a kind of computer readable storage medium,
Computer program is stored on the computer readable storage medium, which performs above-mentioned debt-credit when being run by processor
The step of Forecasting Methodology.
Specifically, which can be general storage medium, such as mobile disk, hard disk, on the storage medium
Computer program when being run, above-mentioned debt-credit Forecasting Methodology is able to carry out, so as to solve to predict borrowing for user in the prior art
The problem of accuracy is relatively low is borrowed, improves the accuracy for carrying out debt-credit prediction, is convenient for accurately finance marketing.
In embodiment provided herein, it should be understood that disclosed system and method, it can be by others side
Formula is realized.System embodiment described above is only schematical, for example, the division of the unit, only one kind are patrolled
Volume function divides, and can have other dividing mode in actual implementation, in another example, multiple units or component can combine or can
To be integrated into another system or some features can be ignored or does not perform.Another point, shown or discussed is mutual
Coupling, direct-coupling or communication connection can be INDIRECT COUPLING or communication link by some communication interfaces, system or unit
It connects, can be electrical, machinery or other forms.
The unit illustrated as separating component may or may not be physically separate, be shown as unit
The component shown may or may not be physical unit, you can be located at a place or can also be distributed to multiple
In network element.Some or all of unit therein can be selected according to the actual needs to realize the mesh of this embodiment scheme
's.
In addition, each functional unit in the embodiment provided in the application can be integrated in a processing unit, also may be used
To be that each unit is individually physically present, can also two or more units integrate in a unit.
If the function is realized in the form of SFU software functional unit and is independent product sale or in use, can be with
It is stored in a computer read/write memory medium.Based on such understanding, the technical solution of the application is substantially in other words
The part contribute to the prior art or the part of the technical solution can be embodied in the form of software product, the meter
Calculation machine software product is stored in a storage medium, is used including some instructions so that a computer equipment (can be
People's computer, server or network equipment etc.) perform each embodiment the method for the application all or part of step.
And aforementioned storage medium includes：USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited
The various media that can store program code such as reservoir (RAM, Random Access Memory), magnetic disc or CD.
It should be noted that：Similar label and letter represents similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined in a attached drawing, does not then need to that it is further defined and explained in subsequent attached drawing, in addition, term " the
One ", " second ", " third " etc. are only used for distinguishing description, and it is not intended that instruction or hint relative importance.
Finally it should be noted that：The specific embodiment of embodiment described above, only the application, to illustrate the application
Technical solution, rather than its limitations, the protection domain of the application is not limited thereto, although with reference to the foregoing embodiments to this Shen
It please be described in detail, it will be understood by those of ordinary skill in the art that：Any one skilled in the art
In the technical scope disclosed in the application, it can still modify to the technical solution recorded in previous embodiment or can be light
It is readily conceivable that variation or equivalent replacement is carried out to which part technical characteristic；And these modifications, variation or replacement, do not make
The essence of corresponding technical solution is detached from the spirit and scope of the embodiment of the present application technical solution.The protection in the application should all be covered
Within the scope of.Therefore, the protection domain of the application described should be subject to the protection scope in claims.
Claims (10)
1. a kind of training method for borrowing or lending money prediction model, which is characterized in that this method includes：
The history debt-credit data of multiple users based on acquisition, determine debt-credit attributive character and lend-borrow action feature；
Using the debt-credit attributive character as independent variable, using the lend-borrow action feature as dependent variable, structure at least two is in advance
If model is simultaneously trained；
Based on the prediction result that at least two preset model obtains, using preset model fusion method to described at least two
Preset model carries out fusion treatment, obtains debt-credit prediction model.
2. the method as described in claim 1, which is characterized in that the prediction knot obtained based at least two preset model
Fruit carries out fusion treatment at least two preset model using preset model fusion method, specifically includes：
Using the prediction result of at least two preset model as independent variable, using lend-borrow action feature as dependent variable, structure
Fusion Model is simultaneously trained.
3. the method as described in claim 1, which is characterized in that using the debt-credit attributive character as independent variable, borrowed by described in
Behavioural characteristic is borrowed as dependent variable, at least two preset models is built and is trained, specifically include：
The history debt-credit data of multiple users based on acquisition, determine that each user corresponds to the characteristic value of debt-credit attributive character, with
And the characteristic value of corresponding lend-borrow action feature；
Determine at least two preset models；And
For each preset model, each user is corresponded to value of the characteristic value as independent variable of debt-credit attributive character, by correspondence
Value of the characteristic value of lend-borrow action feature as dependent variable, is trained the preset model, obtains completing at least the two of training
A preset model.
4. method as claimed in claim 3, which is characterized in that at least two preset model includes：Neural network prediction
Model；
For each preset model, each user is corresponded to value of the characteristic value as independent variable of debt-credit attributive character, by correspondence
Value of the characteristic value of lend-borrow action feature as dependent variable, is trained the preset model, specifically includes：
For multiple neural network models, following training operation is performed respectively, wherein, multiple neural networks have different nerves
The network number of plies：
Each user is corresponded to value of the characteristic value as independent variable of debt-credit attributive character, by the feature of corresponding lend-borrow action feature
It is worth the value as dependent variable, Current Situation of Neural Network model is trained, Current Situation of Neural Network model parameter is obtained and is used for
Weigh the index value of the pre-set level of model prediction accuracy；
The corresponding neural network model of the highest index value of forecasting accuracy will be characterized as finally determining neural network prediction
Model.
5. method as claimed in claim 3, which is characterized in that the history debt-credit data of the multiple user include the multiple
The history debt-credit number of the history debt-credit data of the first historical time section of user and the second historical time section of the multiple user
According to, and the first historical time section is earlier than the second historical time section；
The history debt-credit data of multiple users based on acquisition, determine that each user corresponds to the characteristic value of debt-credit attributive character, with
And the characteristic value of corresponding lend-borrow action feature, it specifically includes：
The history debt-credit data of the first historical time section of the multiple user based on acquisition, determine that each user corresponds to debt-credit
The characteristic value of attributive character；
The history debt-credit data of the second historical time section of the multiple user based on acquisition, determine that each user corresponds to debt-credit
The characteristic value of behavioural characteristic.
6. a kind of debt-credit Forecasting Methodology, which is characterized in that this method includes：
The history debt-credit data of user to be predicted based on acquisition determine the feature of the debt-credit attributive character of the user to be predicted
Value；
Using the characteristic value of determining debt-credit attributive character as the value of independent variable, input as any one of claim 1-5 is determined
Prediction model is borrowed or lent money, predicts the debt-credit probability of the user to be predicted.
7. a kind of training device for borrowing or lending money prediction model, which is characterized in that the device includes：
First processing module borrows or lends money data for the history of multiple users based on acquisition, determines debt-credit attributive character and borrow
Borrow behavioural characteristic；
Training module, for using it is described debt-credit attributive character as independent variable, using the lend-borrow action feature as dependent variable, structure
It builds at least two preset models and is trained；
Second processing module for the prediction result obtained based at least two preset model, is merged using preset model
Method carries out fusion treatment at least two preset model, obtains debt-credit prediction model.
8. a kind of debt-credit prediction meanss, which is characterized in that the device includes：
Processing module borrows or lends money data for the history of the user to be predicted based on acquisition, determines the debt-credit of the user to be predicted
The characteristic value of attributive character；
Prediction module, the characteristic value of debt-credit attributive character for that will determine are inputted as independent variable as claim 7 determines
Prediction model is borrowed or lent money, predicts the debt-credit probability of the user to be predicted.
9. a kind of computer equipment includes memory, processor and is stored on the memory and can transport on the processor
Capable computer program, which is characterized in that the processor realizes the claims 1 to 5 when performing the computer program
The step of any one of them method.
10. a kind of computer equipment includes memory, processor and is stored on the memory and can be on the processor
The computer program of operation, which is characterized in that the processor realizes 6 institute of the claims when performing the computer program
The step of method stated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810098233.8A CN108182634A (en) | 2018-01-31 | 2018-01-31 | A kind of training method for borrowing or lending money prediction model, debt-credit Forecasting Methodology and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810098233.8A CN108182634A (en) | 2018-01-31 | 2018-01-31 | A kind of training method for borrowing or lending money prediction model, debt-credit Forecasting Methodology and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108182634A true CN108182634A (en) | 2018-06-19 |
Family
ID=62551890
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810098233.8A Pending CN108182634A (en) | 2018-01-31 | 2018-01-31 | A kind of training method for borrowing or lending money prediction model, debt-credit Forecasting Methodology and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108182634A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109102396A (en) * | 2018-08-17 | 2018-12-28 | 北京玖富普惠信息技术有限公司 | A kind of user credit ranking method, computer equipment and readable medium |
| CN109165840A (en) * | 2018-08-20 | 2019-01-08 | 平安科技（深圳）有限公司 | Risk profile processing method, device, computer equipment and medium |
| CN109255506A (en) * | 2018-11-22 | 2019-01-22 | 重庆邮电大学 | A kind of internet finance user's overdue loan prediction technique based on big data |
| CN109359762A (en) * | 2018-08-23 | 2019-02-19 | 阿里巴巴集团控股有限公司 | Risk forecast model generation method, Risk Forecast Method, device and server |
| CN109408583A (en) * | 2018-09-25 | 2019-03-01 | 平安科技（深圳）有限公司 | Data processing method and device, computer readable storage medium, electronic equipment |
| CN109800926A (en) * | 2019-03-04 | 2019-05-24 | 苏州大学 | A kind of line of credit prediction technique, device and medium based on depth integration |
| CN110009384A (en) * | 2019-01-07 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Predict the method and device of operational indicator |
| CN110993118A (en) * | 2020-02-29 | 2020-04-10 | 同盾控股有限公司 | Epidemic situation prediction method, device, equipment and medium based on ensemble learning model |
| CN111105265A (en) * | 2019-10-12 | 2020-05-05 | 平安银行股份有限公司 | Prediction method and device based on customer information, computer equipment and storage medium |
| CN112016799A (en) * | 2020-07-15 | 2020-12-01 | 北京淇瑀信息科技有限公司 | Resource quota allocation method and device and electronic equipment |
| CN112232950A (en) * | 2020-12-10 | 2021-01-15 | 银联商务股份有限公司 | Loan risk assessment method and device, equipment and computer-readable storage medium |
| CN112541817A (en) * | 2020-12-22 | 2021-03-23 | 建信金融科技有限责任公司 | Marketing response processing method and system for potential customers of personal consumption loan |
| CN113129028A (en) * | 2020-01-10 | 2021-07-16 | 联洋国融(北京)科技有限公司 | Rogue user detection system based on time sequence neural network model |
| CN113935824A (en) * | 2021-10-13 | 2022-01-14 | 百融云创科技股份有限公司 | Loan intention estimation method and system based on deep learning model |
-
2018
- 2018-01-31 CN CN201810098233.8A patent/CN108182634A/en active Pending
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109102396A (en) * | 2018-08-17 | 2018-12-28 | 北京玖富普惠信息技术有限公司 | A kind of user credit ranking method, computer equipment and readable medium |
| CN109165840A (en) * | 2018-08-20 | 2019-01-08 | 平安科技（深圳）有限公司 | Risk profile processing method, device, computer equipment and medium |
| CN109165840B (en) * | 2018-08-20 | 2022-06-21 | 平安科技（深圳）有限公司 | Risk prediction processing method, risk prediction processing device, computer equipment and medium |
| WO2020037942A1 (en) * | 2018-08-20 | 2020-02-27 | 平安科技（深圳）有限公司 | Risk prediction processing method and apparatus, computer device and medium |
| CN109359762A (en) * | 2018-08-23 | 2019-02-19 | 阿里巴巴集团控股有限公司 | Risk forecast model generation method, Risk Forecast Method, device and server |
| CN109359762B (en) * | 2018-08-23 | 2021-07-20 | 创新先进技术有限公司 | Risk prediction model generation method, risk prediction device and server |
| CN109408583B (en) * | 2018-09-25 | 2023-04-07 | 平安科技（深圳）有限公司 | Data processing method and device, computer readable storage medium and electronic equipment |
| CN109408583A (en) * | 2018-09-25 | 2019-03-01 | 平安科技（深圳）有限公司 | Data processing method and device, computer readable storage medium, electronic equipment |
| CN109255506A (en) * | 2018-11-22 | 2019-01-22 | 重庆邮电大学 | A kind of internet finance user's overdue loan prediction technique based on big data |
| CN110009384A (en) * | 2019-01-07 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Predict the method and device of operational indicator |
| CN109800926B (en) * | 2019-03-04 | 2023-06-16 | 苏州大学 | Credit line prediction method, device and medium based on depth fusion |
| CN109800926A (en) * | 2019-03-04 | 2019-05-24 | 苏州大学 | A kind of line of credit prediction technique, device and medium based on depth integration |
| CN111105265A (en) * | 2019-10-12 | 2020-05-05 | 平安银行股份有限公司 | Prediction method and device based on customer information, computer equipment and storage medium |
| CN113129028A (en) * | 2020-01-10 | 2021-07-16 | 联洋国融(北京)科技有限公司 | Rogue user detection system based on time sequence neural network model |
| CN110993118A (en) * | 2020-02-29 | 2020-04-10 | 同盾控股有限公司 | Epidemic situation prediction method, device, equipment and medium based on ensemble learning model |
| CN112016799A (en) * | 2020-07-15 | 2020-12-01 | 北京淇瑀信息科技有限公司 | Resource quota allocation method and device and electronic equipment |
| CN112232950A (en) * | 2020-12-10 | 2021-01-15 | 银联商务股份有限公司 | Loan risk assessment method and device, equipment and computer-readable storage medium |
| CN112541817A (en) * | 2020-12-22 | 2021-03-23 | 建信金融科技有限责任公司 | Marketing response processing method and system for potential customers of personal consumption loan |
| CN113935824A (en) * | 2021-10-13 | 2022-01-14 | 百融云创科技股份有限公司 | Loan intention estimation method and system based on deep learning model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108182634A (en) | A kind of training method for borrowing or lending money prediction model, debt-credit Forecasting Methodology and device | |
| CN109800885A (en) | It is determined for the rule of black box machine learning model | |
| CN108416669A (en) | User behavior data processing method, device, electronic equipment and computer-readable medium | |
| CN110147882B (en) | Neural network model training method, crowd diffusion method, device and equipment | |
| CN103927615B (en) | Entity is associated with classification | |
| CN107330741A (en) | Graded electron-like certificate uses Forecasting Methodology, device and electronic equipment | |
| CN108961032A (en) | Borrow or lend money processing method, device and server | |
| CN108053314A (en) | A kind of Loan Demand Forecasting Methodology | |
| CN107507036A (en) | The method and terminal of a kind of data prediction | |
| CN113609345B (en) | Target object association method and device, computing equipment and storage medium | |
| Aggarwal | Attitudinal Choquet integrals and applications in decision making | |
| CN108596393A (en) | A kind of training method and device of Electronic Finance resources model | |
| CN113191838A (en) | Shopping recommendation method and system based on heterogeneous graph neural network | |
| CN108596276A (en) | The naive Bayesian microblog users sorting technique of feature based weighting | |
| CN108596765A (en) | A kind of Electronic Finance resource recommendation method and device | |
| CN109670927A (en) | The method of adjustment and its device of credit line, equipment, storage medium | |
| CN109727125A (en) | Borrowing balance prediction technique, device, server, storage medium | |
| CN106575418A (en) | Suggested keywords | |
| Peng et al. | Intelligent Optimization Model of Enterprise Financial Account Receivable Management | |
| Petchhan et al. | Toward Project Success Forecasting in Reward-based Crowdfunding through Wide-and-Deep Computational Learning | |
| CN115880698B (en) | Depression emotion recognition method based on microblog posting content and social behavior characteristics | |
| US20230351415A1 (en) | Using machine learning to leverage interactions to generate hyperpersonalized actions | |
| US20230351193A1 (en) | Training artificial intelligence engine to identify commonalities among nodes to predict actions | |
| US20230351424A1 (en) | Machine learning algorithm for predicting events using multiple disparate inputs | |
| US11989777B2 (en) | Pairing and grouping user profiles accessed from pre-current systems |
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 | ||
| CB02 | Change of applicant information |
Address after: 101-8, 1st floor, building 31, area 1, 188 South Fourth Ring Road West, Fengtai District, Beijing Applicant after: Guoxin Youyi Data Co., Ltd Address before: 100070, No. 188, building 31, headquarters square, South Fourth Ring Road West, Fengtai District, Beijing Applicant before: SIC YOUE DATA Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180619 |
|
| RJ01 | Rejection of invention patent application after publication |