CN106408325A - User consumption behavior prediction analysis method based on user payment information and system - Google Patents
User consumption behavior prediction analysis method based on user payment information and system Download PDFInfo
- Publication number
- CN106408325A CN106408325A CN201610754174.6A CN201610754174A CN106408325A CN 106408325 A CN106408325 A CN 106408325A CN 201610754174 A CN201610754174 A CN 201610754174A CN 106408325 A CN106408325 A CN 106408325A
- Authority
- CN
- China
- Prior art keywords
- feature
- user
- loss
- variable
- subsystem
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
Abstract
The invention belongs to the Internet payment technology field and provides a user consumption behavior prediction analysis method based on the user payment information and a system. The system comprises a data acquiring subsystem, a data gathering subsystem, a standard characteristic formation subsystem, a prediction model determining subsystem and an actual test subsystem, wherein the data acquiring subsystem is used for acquiring the transaction data, the behavior data and the customer service feedback data of a user, the data gathering subsystem is used for forming characteristics and characteristic variables corresponding to the characteristics, the prediction model determining subsystem is used for determining a user loss prediction model consistent with stability standards, and the actual test subsystem is used for employing the user loss prediction model consistent with the stability standards to process characteristic variables corresponding to standard characteristics to acquire a user consumption behavior prediction result. Through the method and the system, the bottom-level Internet payment information can be mined and utilized to a maximum degree, and user consumption behaviors can be effectively predicted.
Description
Technical field
The present invention relates to internet payment technical field is and in particular to a kind of customer consumption row based on user's payment information
For prediction analysis method and system.
Background technology
At present, digital entertainment, ecommerce are rapidly growing, and network payment technology is also fast-developing.During network payment
Substantial amounts of user's payment information can be produced, it has from the payment the most basic and direct payment data of user.
Meanwhile, each businessman or trade company are also required to obtain the demand of user, or even launch questionnaire survey, are needed with capturing user
Ask, adjust migration efficiency.Part businessman can analyze user's request or turnover rate, but analysis result be accurate in conjunction with transaction data
Degree is poor.
How to excavate and to apply payment data to greatest extent, effectively predict consumer consumption behavior, be people in the art
The problem of member's urgent need to resolve.
Content of the invention
For defect of the prior art, the present invention provides a kind of prediction of the consumer consumption behavior based on user's payment information
Analysis method and system, can excavate and apply payment data to greatest extent, effectively predict consumer consumption behavior.
In a first aspect, the present invention provides a kind of consumer consumption behavior prediction analysis method based on user's payment information, should
Method includes：
Data acquisition step：Obtain transaction data, behavioral data and the customer service feedback data of user；
Data summarization step：Collect transaction data, behavioral data and customer service feedback data, form feature corresponding with this feature
Characteristic variable；
Standard feature forming step：According to the corresponding professional knowledge of feature, reject feature corresponding off-note variable；Root
According to characteristic variable, construct and screen complex variable；The feature meeting linear requirements in feature is screened using lasso trick algorithm, forms mark
Quasi- feature；
Forecast model determines step：The sample set of pre-acquiring is divided into by training set and test set according to setting ratio；Adopt
Use customer loss forecast model, training set is predicted, acquisition predicts the outcome；According to the actual stream predicting the outcome with training set
Lose-submission label, obtain Training Quality Index；Using customer loss forecast model, process test collection, obtain test quality index；According to
Training Quality Index and test quality index, are determined for compliance with the customer loss forecast model of stability criterion；
Actual test step：According to standard feature, obtain the corresponding characteristic variable of standard feature of user to be tested；Using
Stability standard compliant customer loss forecast model, processes the corresponding characteristic variable of standard feature, obtains consumer consumption behavior
Predict the outcome.
Further, according to characteristic variable, construct and screen complex variable, specifically include：
According to characteristic variable, square complex variable of structural features variable and/or intersect complex variable,
According to the business meaning of characteristic variable and complex variable, screening meets the complex variable of business demand.
Based on the consumer consumption behavior prediction analysis method embodiment of above-mentioned any user's payment information, further, exist
After constructing and screening complex variable, screen the feature meeting linear requirements in feature using lasso trick algorithm before, the method is also
Including：Variance inflation factor testing result according to feature and/or the business meaning of feature, whether judging characteristic meets linearly will
Ask.
Further, the variance inflation factor testing result according to feature, whether judging characteristic meets linear requirements, specifically
Including：
If the variance inflation factor testing result of feature is less than linear threshold, judging characteristic meets linear requirements；If special
The variance inflation factor testing result levied is more than or equal to linear threshold, then judging characteristic does not meet linear requirements, and linear threshold is
Set in advance.
Based on the consumer consumption behavior prediction analysis method embodiment of above-mentioned any user's payment information, further, adopt
With stability standard compliant customer loss forecast model, process the corresponding characteristic variable of standard feature, obtain customer consumption row
For predicting the outcome, specifically include：
Using one week loss model of the standard compliant user of stability, process the corresponding characteristic variable of standard feature, obtain
Loss probability；
If loss probability is more than or equal to loss threshold value, judge this user as by user of running off；
If loss probability is less than loss threshold value, judge this user as user of not running off；
Stability standard compliant customer loss forecast model includes one week loss model of the standard compliant user of stability,
Consumer consumption behavior predicts the outcome including by loss user and user of not running off.
Second aspect, the present invention provides a kind of consumer consumption behavior hypothesis analysis system based on user's payment information, should
System includes data acquisition subsystem, data summarization subsystem, standard feature formed subsystem, forecast model determine subsystem and
Actual test subsystem.Data acquisition subsystem is used for obtaining transaction data, behavioral data and the customer service feedback data of user；Number
It is used for collecting transaction data, behavioral data and customer service feedback data according to collecting subsystem, form feature and the corresponding spy of this feature
Levy variable；Standard feature forms subsystem and is used for according to the corresponding professional knowledge of feature, rejects the corresponding off-note of feature and becomes
Amount, is additionally operable to according to characteristic variable, constructs and screen complex variable, and for meeting line using in lasso trick algorithm screening feature
Property require feature, formed standard feature；Forecast model determines that subsystem is used for the sample set according to setting ratio by pre-acquiring
It is divided into training set and test set, using customer loss forecast model, training set is predicted, acquisition predicts the outcome, according to
The actual loss label predicting the outcome with training set, obtains Training Quality Index, using customer loss forecast model, process test
Collection, obtains test quality index, and for according to Training Quality Index and test quality index, being determined for compliance with stability criterion
Customer loss forecast model；Actual test subsystem is used for obtaining the standard feature pair of user to be tested according to standard feature
The characteristic variable answered, is additionally operable to, using stability standard compliant customer loss forecast model, process the corresponding spy of standard feature
Levy variable, obtain consumer consumption behavior and predict the outcome.
Further, standard feature forms subsystem and according to characteristic variable, is constructing and screening complex variable, concrete use
In：According to characteristic variable, square complex variable of structural features variable and/or intersect complex variable, according to characteristic variable and multiple
Close the business meaning of variable, screening meets the complex variable of business demand.
Based on the consumer consumption behavior hypothesis analysis system embodiment of above-mentioned any user's payment information, further, it is somebody's turn to do
System also includes synteny and judges subsystem, for according to the variance inflation factor testing result of feature and/or the business of feature
Meaning, whether judging characteristic meets linear requirements.
Further, synteny judges subsystem in the variance inflation factor testing result according to feature, and judging characteristic is
Not no when meeting linear requirements, specifically for：If the variance inflation factor testing result of feature is less than linear threshold, judging characteristic
Meet linear requirements, if the variance inflation factor testing result of feature is more than or equal to linear threshold, judging characteristic is not inconsistent zygonema
Property require, linear threshold be set in advance.
Based on the consumer consumption behavior hypothesis analysis system embodiment of above-mentioned any user's payment information, further, real
Border test subsystems specifically for：Using one week loss model of the standard compliant user of stability, process standard feature corresponding
Characteristic variable, obtains loss probability, if loss probability is more than or equal to loss threshold value, judges this user as by user of running off, if
Loss probability is less than loss threshold value, then judge this user as user of not running off, and the standard compliant customer loss of stability predicts mould
Type includes one week loss model of the standard compliant user of stability, and consumer consumption behavior predicts the outcome including by loss user and not
Loss user.
As shown from the above technical solution, the consumer consumption behavior prediction analysis method based on user's payment information for the present invention and
System, can be in conjunction with bottom data such as traditional transaction data, the behavioral data of user and customer service feedback data, to user's
Behavior carry out omnibearing portray, user's real conditions can be reflected.
Meanwhile, the consumer consumption behavior prediction analysis method based on user's payment information for the present invention and system are capable of determining that
Meet the standard feature of linear requirements, and meet the customer loss forecast model of stability criterion, pre- using this customer loss
Survey the corresponding characteristic variable of models treated mark feature, the behaviors such as customer loss precisely can be predicted, to support trade company
Operation and precision marketing etc..
Therefore, the consumer consumption behavior prediction analysis method based on user's payment information for the present embodiment and system, can be
The excavation of limits and application bottom internet payment information, effectively predict consumer consumption behavior, are that trade company is precisely sought
Pin provides accurate Informational support.
Brief description
In order to be illustrated more clearly that the specific embodiment of the invention or technical scheme of the prior art, below will be to concrete
In embodiment or description of the prior art, the accompanying drawing of required use is briefly described.In all of the figs, similar element
Or partly typically identified by similar reference.In accompanying drawing, each element or part might not be drawn according to actual ratio.
Fig. 1 shows a kind of consumer consumption behavior prediction analysis method based on user's payment information provided by the present invention
Flow chart；
Fig. 2 shows a kind of consumer consumption behavior hypothesis analysis system based on user's payment information provided by the present invention
Structural representation.
Specific embodiment
Below in conjunction with accompanying drawing, the embodiment of technical solution of the present invention is described in detail.Following examples are only used for
Clearly technical scheme is described, is therefore intended only as example, and the protection of the present invention can not be limited with this
Scope.
It should be noted that unless otherwise stated, technical term used in this application or scientific terminology should be this
The ordinary meaning that bright one of ordinary skill in the art are understood.
In a first aspect, the embodiment of the present invention provides a kind of consumer consumption behavior forecast analysis side based on user's payment information
Method, in conjunction with Fig. 1, the method includes：
Data acquisition step S1：Obtain transaction data, behavioral data and the customer service feedback data of user；
In actual application, the method can by client gather user in cashier within behavioral data,
Application message or other information, or transaction data is obtained by basic platform.Wherein, behavioral data is mainly the behaviour showing user
Make, such as enter game initialization, adjusted cashier, show cashier, adjusted the means of payment, the click of the various means of payment, tune
With and return, cashier click on supplement with money, the various means of payment are supplemented with money, pay and are returned, change payment cipher, log in, registration etc.
Deng；Application message is mainly current application information, channel information, payment interface version, application version, applies bag name, terminal system
System version, terminal currently mounted all application messages；Can adopt when user's point is opened cashier or paid
Collect the payment data of user, including but not limited to purchase commodity, payment, time of payment, the means of payment；Whenever trade company connects
Enter to like the payment interface of shellfish, their merchant information and application message are also the key element analyzed.In addition, love shellfish is equipped with for user
The customer service of 7*24 hour, can collect the calling information of user, reimbursement information etc. from customer service there.
Data summarization step S2：Collect transaction data, behavioral data and customer service feedback data, form feature and this feature pair
The characteristic variable answered；
Wherein, this step can be processed to the data of collection with reference to data warehouse framework using big data technology and be converged
Always, specifically utilize hadoop and hive to be platform, data is processed, first loads the data source in all data to data warehouses
Layer, the wide top layer built up then in conjunction with merchant information, application message and other information group, generate business finally according to service needed
Layer table.The conversion of transaction data, behavioral data and customer service feedback data is characterized the corresponding characteristic variable with this feature, after being easy to
The transferring or processing of continuous data.
Standard feature forming step S3：
According to the corresponding professional knowledge of feature, reject feature corresponding off-note variable；
Wherein, off-note variable refer mainly to certain feature characteristic variable deviate average too remote, they would generally make mould
Type produces and significantly deviates, thus affecting model quality, needs to reject.Generally require and come really with reference to professional knowledge and statistical data
Fixed abnormal standard, the exceptional value standard of different characteristic has difference；
According to characteristic variable, construct and screen complex variable；
The feature meeting linear requirements in feature is screened using lasso trick algorithm, forms standard feature；
Wherein, lasso trick (Lasso) algorithm can solve the problem that the conllinear sex chromosome mosaicism of feature, and can automatically select feature.Here,
The standard feature being obtained using lasso trick (Lasso) algorithm is as follows：
PAY_MONEY_2WEEKS_LOG (logarithm of nearly two weeks total transaction amounts)
PAY_MONEY_4WEEKS_LOG (nearly surrounding total transaction amount is taken the logarithm)
PAY_MONEY_1WEEK_RATIO (nearly one week total transaction amount and week before last total transaction amount ratio)
PAY_MONEY_2WEEKS_RATIO (nearly two weeks total transaction amounts and last two weeks total transaction amount ratio)
PAY_TIMES_HISTORY (historical trading number of times)
PAY_TIMES_1WEEK (nearly one week transaction count)
PAY_TIMES_2WEEKS (nearly two weeks transaction counts)
PAY_TIMES_4WEEKS (nearly 4 weeks transaction counts)
PAY_TIMES_WORKDAY_RATIO_1WEEK (nearly one week weekend transaction natural law and work Day Trading natural law ratio)
PAY_TIMES_WORKDAY_RATIO_2WEEKS (nearly two week weekend transaction natural law and work Day Trading natural law ratio
Value)
PAY_TIMES_WORKDAY_RATIO_4WEEKS (nearly surrounding weekend transaction natural law and work Day Trading natural law ratio
Value)
(there is within nearly one week the natural law of transaction) in PAY_DAYS_1WEEK
(there is within nearly two weeks the natural law of transaction) in PAY_DAYS_2WEEKS
PAY_DAYS_4WEEKS (nearly surrounding has the natural law of transaction)
(there is within nearly one week transaction natural law and there is transaction natural law ratio with week before last) in PAY_DAYS_RATIO_1WEEK
PAY_SILENCE_1WEEK (whether silence natural law is more than one week)
PAY_FIRSTPAY_IS_NEW (whether initial transaction)
PAY_D_ENTROPY (same day is poor with the entropy of the previous day)
PAY_D_TIMEDEV_EUCLIDEANDISTANCE (same day is with the vectorial Euclidean distance of the previous day)
Forecast model determines step S4：
The sample set of pre-acquiring is divided into by training set and test set according to setting ratio；
Here, dividing test set, training set two parts by sample set, set up model by training set, model will after building up
It is used in tests on test set.Rational model should closely otherwise just have excessively in the performance of training set and test set
The problem of matching, namely model is fine to the prediction of training set, but the prediction to test set is very poor.If there is significantly excessive
Fitting problems, then model will lose the meaning of popularization and application.In dividing training set and test set operation, randomly draw 80% sample
This collection does training set, and remaining sample does test set.
Using customer loss forecast model, training set is predicted, acquisition predicts the outcome；
The algorithm that customer loss forecast model adopts includes logistic regression algorithm, and the application conditions of this algorithm are each features
Between not significantly conllinear sex chromosome mosaicism.The complexity that this algorithm calculates is not high, and should be readily appreciated that and realize.
According to the actual loss label predicting the outcome with training set, obtain Training Quality Index；
Wherein, the primary quality measure of evaluation model quality includes overall accuracy rate, recall rate, precision, F1-Measure,
This four quality index have internal relation, and closer to 1, then the prediction effect of this model is better for single quality index.Here, the party
Method can only provide the result of overall accuracy rate index, and overall accuracy rate namely the correct characteristic variable sum of prediction account for always special horizontal stroke
Scaling of variables, for the model that we are set up, the overall rate of accuracy reached of training set, to two asterisks, thinks effect at once
Well.
Using customer loss forecast model, process test collection, obtain test quality index；
According to Training Quality Index and test quality index, it is determined for compliance with the customer loss forecast model of stability criterion；
Here, this model is from the point of view of the performance on test set, give a forecast the aforementioned four quality index obtaining to test set
With the corresponding quality index of training set closely, result is balance between training set and test set, illustrates that this model has
There is good stability.
Actual test step S5：
According to standard feature, obtain the corresponding characteristic variable of standard feature of user to be tested；
Using stability standard compliant customer loss forecast model, process the corresponding characteristic variable of standard feature, obtain
Consumer consumption behavior predicts the outcome.
As shown from the above technical solution, the consumer consumption behavior forecast analysis side based on user's payment information for the present embodiment
Method, can be in conjunction with the bottom data such as traditional transaction data, the behavioral data of user and customer service feedback data, the row to user
For carry out omnibearing portray, user's real conditions can be reflected.
Meanwhile, the method is capable of determining that the standard feature meeting linear requirements, and the user meeting stability criterion
Attrition prediction model, processes the mark corresponding characteristic variable of feature using this customer loss forecast model, can be to customer loss etc.
Behavior is precisely predicted, to support operation and precision marketing of trade company etc..Here, can be shown at analysis using webpage various dimensions
Predicting the outcome after reason, including but not limited to sales list, pay custom, paying permeability, game playing habits etc..
Therefore, the consumer consumption behavior prediction analysis method based on user's payment information for the present embodiment, can be to greatest extent
Excavation with application bottom internet payment information, effectively predict consumer consumption behavior, carry out precision marketing offer for trade company
Accurately Informational support.
Specifically, in standard feature forming process, in order to be combined further to above-mentioned data, more to excavate
Information, thus improving the quality of follow-up modeling, according to characteristic variable, constructing and screening complex variable, specifically including：According to feature
Variable, square complex variable of structural features variable and/or intersection complex variable, according to the business of characteristic variable and complex variable
Meaning, screening meets the complex variable of business demand.Characteristic variable can be carried out itself and be multiplied by the method, or special with others
Levy variable and carry out multiplication cross, obtain substantial amounts of complex variable, further according to the important level of characteristic variable itself, and compound change
The business meaning of amount, filters out the complex variable meeting business demand.
After constructing and screening complex variable, using lasso trick algorithm screen feature in meet linear requirements feature it
Before, the method also includes：Variance inflation factor testing result according to feature and/or the business meaning of feature, judging characteristic is
No meet linear requirements.If the variance inflation factor testing result of feature is less than linear threshold, judging characteristic meets linearly will
Ask；If the variance inflation factor testing result of feature is more than or equal to linear threshold, judging characteristic does not meet linear requirements, linearly
Threshold value is set in advance.Feature is checked using variance inflation factor (Variance Inflation Factor, VIF)
Conllinear sex chromosome mosaicism.It has been generally acknowledged that VIF>10 problems that there is synteny, also will judge according to service logic sometimes.With VIF>
10 is standard, finds that the conllinear sex chromosome mosaicism of feature is obvious after being computed.Business meaning according to variance inflation factor or feature
Justice, whether judging characteristic has conllinear sex chromosome mosaicism, further filters out, using lasso trick algorithm, the feature meeting linear requirements, with
Improve data-handling efficiency and accuracy.
Specifically, during actual test, for the application of certain class, first determine the ID needing to give a forecast, and
Obtain corresponding 30 characteristic variables, such as historical trading value data, historical trading frequency data, transaction silence natural law data,
Dealing money behavioral pattern variable, exchange hour behavioral pattern variable, time difference behavioral pattern variable in the customer transaction time period,
The corresponding feature of these characteristic variables, that is, above-mentioned 19 standard features.For user's one week attrition prediction of future, will be above-mentioned
Characteristic variable inputs one week loss model of user, and this model can export corresponding loss Probability p, and probability being classified as more than 0.5 will
Loss user, the user that is classified as not running off less than or equal to 0.5；Equally, for user's two weeks attrition prediction of future, by above-mentioned spy
Levy variable input two weeks loss models of user, and be classified as running off or will not running off by user by same standard.
Second aspect, the embodiment of the present invention provides a kind of consumer consumption behavior forecast analysis system based on user's payment information
System, in conjunction with Fig. 2, this system includes data acquisition subsystem 1, data summarization subsystem 2, standard feature formation subsystem 3, prediction
Model determines subsystem 4 and actual test subsystem 5.Data acquisition subsystem 1 is used for obtaining transaction data, the behavior number of user
According to customer service feedback data.Data summarization subsystem 2 is used for collecting transaction data, behavioral data and customer service feedback data, is formed
Feature and the corresponding characteristic variable of this feature.Standard feature forms subsystem 3 and is used for, according to the corresponding professional knowledge of feature, rejecting
Feature corresponding off-note variable, is additionally operable to according to characteristic variable, constructs and screen complex variable, and be used for adopting lasso trick
Algorithm screens the feature meeting linear requirements in feature, forms standard feature.Forecast model determines subsystem 4 for according to setting
The sample set of pre-acquiring is divided into training set and test set by ratio, using customer loss forecast model, training set is carried out pre-
Survey, acquisition predicts the outcome, according to the actual loss label predicting the outcome with training set, obtain Training Quality Index, using user
Attrition prediction model, process test collection, obtain test quality index, and for being referred to according to Training Quality Index and test quality
Mark, is determined for compliance with the customer loss forecast model of stability criterion.Actual test subsystem 5, for according to standard feature, obtaining
The corresponding characteristic variable of standard feature of user to be tested, is additionally operable to predict mould using the standard compliant customer loss of stability
Type, processes the corresponding characteristic variable of standard feature, obtains consumer consumption behavior and predicts the outcome.
As shown from the above technical solution, the consumer consumption behavior forecast analysis system based on user's payment information for the present embodiment
System, can be in conjunction with the bottom data such as traditional transaction data, the behavioral data of user and customer service feedback data, the row to user
For carry out omnibearing portray, user's real conditions can be reflected.
Meanwhile, this system is capable of determining that the standard feature meeting linear requirements, and the user meeting stability criterion
Attrition prediction model, processes the mark corresponding characteristic variable of feature using this customer loss forecast model, can be to customer loss etc.
Behavior is precisely predicted, to support operation and precision marketing of trade company etc..Here, can be shown at analysis using webpage various dimensions
Predicting the outcome after reason, including but not limited to sales list, pay custom, paying permeability, game playing habits etc..
Therefore, the consumer consumption behavior hypothesis analysis system based on user's payment information for the present embodiment, can be to greatest extent
Excavation with application bottom internet payment information, effectively predict consumer consumption behavior, carry out precision marketing offer for trade company
Accurately Informational support.
In order to be combined further to above-mentioned data, to excavate more information, thus improving the quality of follow-up modeling,
Standard feature forms subsystem 3 according to characteristic variable, constructing and screen complex variable, specifically for：According to characteristic variable, structure
Make square complex variable of characteristic variable and/or intersect complex variable, according to the business meaning of characteristic variable and complex variable, sieve
Choosing meets the complex variable of business demand.This standard feature forms subsystem 3 and characteristic variable can be carried out itself multiplication, or with
Other characteristic variables carry out multiplication cross, obtain substantial amounts of complex variable, further according to the important level of characteristic variable itself, with
And the business meaning of complex variable, filter out the complex variable meeting business demand.
Meanwhile, the present embodiment also includes synteny based on the consumer consumption behavior hypothesis analysis system of user's payment information and sentences
Disconnected subsystem, it is used for according to the variance inflation factor testing result of feature and/or the business meaning of feature, and whether judging characteristic
Meet linear requirements.Synteny judges subsystem in the variance inflation factor testing result according to feature, and whether judging characteristic accords with
Linear when requiring, specifically for：If the variance inflation factor testing result of feature is less than linear threshold, judging characteristic meets
Linear requirements, if the variance inflation factor testing result of feature is more than or equal to linear threshold, judging characteristic is not inconsistent linear wanting
Ask, linear threshold is set in advance.Synteny judges the business meaning according to variance inflation factor or feature for the subsystem, judges
Whether feature has conllinear sex chromosome mosaicism, further filters out, using lasso trick algorithm, the feature meeting linear requirements, to improve number
According to treatment effeciency and accuracy.
During actual test, if user to be obtained was in the attrition prediction result of following a week, actual test subsystem
System 5 specifically for：Using one week loss model of the standard compliant user of stability, process the corresponding characteristic variable of standard feature,
Obtain loss probability, if loss probability is more than or equal to loss threshold value, judge this user as by user of running off, if loss probability is little
In loss threshold value, then judge this user as user of not running off, stability standard compliant customer loss forecast model includes stable
Property one week loss model of standard compliant user, consumer consumption behavior predicts the outcome including by loss user and user of not running off.
Finally it should be noted that：Various embodiments above only in order to technical scheme to be described, is not intended to limit；To the greatest extent
Pipe has been described in detail to the present invention with reference to foregoing embodiments, it will be understood by those within the art that：Its according to
So the technical scheme described in foregoing embodiments can be modified, or wherein some or all of technical characteristic is entered
Row equivalent；And these modifications or replacement, do not make the essence of appropriate technical solution depart from various embodiments of the present invention technology
The scope of scheme, it all should be covered in the middle of the claim of the present invention and the scope of description.
Claims (10)
1. a kind of consumer consumption behavior prediction analysis method based on user's payment information is it is characterised in that include：
Data acquisition step：
Obtain transaction data, behavioral data and the customer service feedback data of user；
Data summarization step：
Collect described transaction data, described behavioral data and described customer service feedback data, form feature and the corresponding spy of this feature
Levy variable；
Standard feature forming step：
According to the corresponding professional knowledge of described feature, reject described feature corresponding off-note variable；
According to described characteristic variable, construct and screen complex variable；
The feature meeting linear requirements in described feature is screened using lasso trick algorithm, forms standard feature；
Forecast model determines step：
The sample set of pre-acquiring is divided into by training set and test set according to setting ratio；
Using customer loss forecast model, described training set is predicted, acquisition predicts the outcome；
According to the described actual loss label predicting the outcome with described training set, obtain Training Quality Index；
Using described customer loss forecast model, process described test set, obtain test quality index；
According to described Training Quality Index and described test quality index, it is determined for compliance with the customer loss prediction mould of stability criterion
Type；
Actual test step：
According to described standard feature, obtain the corresponding characteristic variable of described standard feature of user to be tested；
Using described stability standard compliant customer loss forecast model, process the corresponding characteristic variable of described standard feature,
Obtain consumer consumption behavior to predict the outcome.
2. according to claim 1 the consumer consumption behavior prediction analysis method based on user's payment information it is characterised in that
Described according to characteristic variable, construct and screen complex variable, specifically include：
According to described characteristic variable, construct square complex variable of described characteristic variable and/or intersect complex variable,
According to the business meaning of described characteristic variable and described complex variable, screening meets the complex variable of business demand.
3. the consumer consumption behavior prediction analysis method based on user's payment information according to claim 1 or claim 2, its feature exists
In,
Described construct and screen complex variable after, using lasso trick algorithm screen feature in meet linear requirements feature it
Before, the method also includes：Variance inflation factor testing result according to described feature and/or the business meaning of described feature, sentence
Whether described feature of breaking meets linear requirements.
4. according to claim 3 the consumer consumption behavior prediction analysis method based on user's payment information it is characterised in that
The described variance inflation factor testing result according to feature, whether judging characteristic meets linear requirements, specifically includes：
If the variance inflation factor testing result of described feature is less than linear threshold, judge the linear requirement of described character symbol；
If the variance inflation factor testing result of described feature is more than or equal to described linear threshold, judge that described feature does not meet
Linear requirements, described linear threshold is set in advance.
5. according to claim 1 the consumer consumption behavior prediction analysis method based on user's payment information it is characterised in that
The standard compliant customer loss forecast model of described employing stability, processes the corresponding characteristic variable of standard feature, obtains
Consumer consumption behavior predicts the outcome, and specifically includes：
Using one week loss model of the standard compliant user of stability, process the corresponding characteristic variable of described standard feature, obtain
Loss probability；
If described loss probability is more than or equal to loss threshold value, judge this user as by user of running off；
If described loss probability is less than described loss threshold value, judge this user as user of not running off；
The standard compliant customer loss forecast model of described stability includes the standard compliant user of described stability and runs off for one week
Model, described consumer consumption behavior predicts the outcome including described by loss user with described user of not running off.
6. a kind of consumer consumption behavior hypothesis analysis system based on user's payment information is it is characterised in that include：
Data acquisition subsystem, for obtaining transaction data, behavioral data and the customer service feedback data of user；
Data summarization subsystem, is used for collecting described transaction data, described behavioral data and described customer service feedback data, is formed special
The corresponding characteristic variable of this feature of seeking peace；
Standard feature forms subsystem, for according to the corresponding professional knowledge of described feature, rejecting the corresponding exception of described feature
Characteristic variable, is additionally operable to, according to described characteristic variable, construct and screen complex variable, and for institute is screened using lasso trick algorithm
State the feature meeting linear requirements in feature, form standard feature；
Forecast model determines subsystem, for the sample set of pre-acquiring is divided into by training set and test set according to setting ratio,
Using customer loss forecast model, described training set is predicted, acquisition predicts the outcome, according to described predict the outcome with described
The actual loss label of training set, obtains Training Quality Index, using described customer loss forecast model, processes described test
Collection, obtains test quality index, and steady for according to described Training Quality Index and described test quality index, being determined for compliance with
The customer loss forecast model of qualitative criteria；
Actual test subsystem, for according to described standard feature, obtaining the corresponding spy of described standard feature of user to be tested
Levy variable, be additionally operable to, using described stability standard compliant customer loss forecast model, process described standard feature corresponding
Characteristic variable, obtains consumer consumption behavior and predicts the outcome.
7. according to claim 6 the consumer consumption behavior hypothesis analysis system based on user's payment information it is characterised in that
Described standard feature forms subsystem according to characteristic variable, constructing and screen complex variable, specifically for：According to described
Characteristic variable, constructs square complex variable of described characteristic variable and/or intersects complex variable, according to described characteristic variable and institute
State the business meaning of complex variable, screening meets the complex variable of business demand.
8. the consumer consumption behavior hypothesis analysis system based on user's payment information according to claim 6 or 7, its feature exists
In this system also includes synteny and judges subsystem, for the variance inflation factor testing result according to described feature and/or institute
State the business meaning of feature, judge whether described feature meets linear requirements.
9. according to claim 8 the consumer consumption behavior hypothesis analysis system based on user's payment information it is characterised in that
Described synteny judges subsystem in the variance inflation factor testing result according to feature, and whether judging characteristic meets linearly
During requirement, specifically for：If the variance inflation factor testing result of described feature is less than linear threshold, judge described character symbol
Linear requirement, if the variance inflation factor testing result of described feature is more than or equal to described linear threshold, judges described spy
Levy and do not meet linear requirements, described linear threshold is set in advance.
10. the consumer consumption behavior hypothesis analysis system based on user's payment information according to claim 6, its feature exists
In,
Described actual test subsystem, specifically for：Using one week loss model of the standard compliant user of stability, process described
The corresponding characteristic variable of standard feature, obtains loss probability, if described loss probability is more than or equal to loss threshold value, judges this use
Family is by user of running off, if described loss probability is less than described loss threshold value, judges this user as user of not running off,
The standard compliant customer loss forecast model of described stability includes the standard compliant user of described stability and runs off for one week
Model, described consumer consumption behavior predicts the outcome including described by loss user with described user of not running off.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610754174.6A CN106408325A (en) | 2016-08-29 | 2016-08-29 | User consumption behavior prediction analysis method based on user payment information and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610754174.6A CN106408325A (en) | 2016-08-29 | 2016-08-29 | User consumption behavior prediction analysis method based on user payment information and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106408325A true CN106408325A (en) | 2017-02-15 |
Family
ID=58003693
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610754174.6A Pending CN106408325A (en) | 2016-08-29 | 2016-08-29 | User consumption behavior prediction analysis method based on user payment information and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106408325A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107688865A (en) * | 2017-07-31 | 2018-02-13 | 上海恺英网络科技有限公司 | Identify the method and apparatus of potential high consumption user in online game |
| CN109492891A (en) * | 2018-10-26 | 2019-03-19 | 阿里巴巴集团控股有限公司 | Customer churn prediction technique and device |
| CN109508329A (en) * | 2018-12-07 | 2019-03-22 | 广州市诚毅科技软件开发有限公司 | Customer churn method for early warning, system and storage medium based on broadcasting and TV big data |
| CN111178987A (en) * | 2020-04-10 | 2020-05-19 | 支付宝(杭州)信息技术有限公司 | Method and device for training user behavior prediction model |
| CN111738775A (en) * | 2020-07-03 | 2020-10-02 | 支付宝(杭州)信息技术有限公司 | Training method and system for user willingness-to-pay prediction model |
| CN111783999A (en) * | 2020-07-01 | 2020-10-16 | 北京知因智慧科技有限公司 | Data processing method and device |
| CN112330047A (en) * | 2020-11-18 | 2021-02-05 | 交通银行股份有限公司 | Credit card repayment probability prediction method based on user behavior characteristics |
-
2016
- 2016-08-29 CN CN201610754174.6A patent/CN106408325A/en active Pending
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107688865A (en) * | 2017-07-31 | 2018-02-13 | 上海恺英网络科技有限公司 | Identify the method and apparatus of potential high consumption user in online game |
| CN109492891A (en) * | 2018-10-26 | 2019-03-19 | 阿里巴巴集团控股有限公司 | Customer churn prediction technique and device |
| CN109492891B (en) * | 2018-10-26 | 2022-04-29 | 创新先进技术有限公司 | User loss prediction method and device |
| CN109508329A (en) * | 2018-12-07 | 2019-03-22 | 广州市诚毅科技软件开发有限公司 | Customer churn method for early warning, system and storage medium based on broadcasting and TV big data |
| CN111178987A (en) * | 2020-04-10 | 2020-05-19 | 支付宝(杭州)信息技术有限公司 | Method and device for training user behavior prediction model |
| CN111178987B (en) * | 2020-04-10 | 2020-06-30 | 支付宝(杭州)信息技术有限公司 | Method and device for training user behavior prediction model |
| CN111783999A (en) * | 2020-07-01 | 2020-10-16 | 北京知因智慧科技有限公司 | Data processing method and device |
| CN111738775A (en) * | 2020-07-03 | 2020-10-02 | 支付宝(杭州)信息技术有限公司 | Training method and system for user willingness-to-pay prediction model |
| CN112330047A (en) * | 2020-11-18 | 2021-02-05 | 交通银行股份有限公司 | Credit card repayment probability prediction method based on user behavior characteristics |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106408325A (en) | User consumption behavior prediction analysis method based on user payment information and system | |
| US11068789B2 (en) | Dynamic model data facility and automated operational model building and usage | |
| US8630892B2 (en) | Churn analysis system | |
| US8874674B2 (en) | System for optimizing social networking | |
| WO2019037202A1 (en) | Method and apparatus for recognising target customer, electronic device and medium | |
| Chen | The gamma CUSUM chart method for online customer churn prediction | |
| US11461343B1 (en) | Prescriptive analytics platform and polarity analysis engine | |
| US20230409906A1 (en) | Machine learning based approach for identification of extremely rare events in high-dimensional space | |
| CN111340606A (en) | Full-process income auditing method and device | |
| Chen et al. | A combined mining-based framework for predicting telecommunications customer payment behaviors | |
| CN113609193A (en) | Method and device for training prediction model for predicting customer transaction behavior | |
| CN115545886A (en) | Overdue risk identification method, overdue risk identification device, overdue risk identification equipment and storage medium | |
| CN107784511A (en) | A kind of customer loss Forecasting Methodology and device | |
| CN113570398A (en) | Promotion data processing method, model training method, system and storage medium | |
| CN109146667B (en) | Method for constructing external interface comprehensive application model based on quantitative statistics | |
| CN112950359A (en) | User identification method and device | |
| CN114493686A (en) | Operation content generation and pushing method and device | |
| CN113674013A (en) | Advertisement bidding adjustment method and system based on merchant self-defined rules | |
| WO2020150597A1 (en) | Systems and methods for entity performance and risk scoring | |
| Guo | Probabilistic forecasting in decision-making: new methods and applications | |
| CN114997879B (en) | Payment routing method, device, equipment and storage medium | |
| Wah | Some applications of data mining | |
| CN117291640A (en) | Method and device for identifying consumption behavior, electronic equipment and storage medium | |
| CN115688998A (en) | House rent income prediction method and device, electronic equipment and storage medium | |
| CN115760196A (en) | Activity recommendation method, device, terminal and program product for long-tail users |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170215 |