CN116843383A - Individualized excitation method and device based on counterfactual identification and estimation - Google Patents

Individualized excitation method and device based on counterfactual identification and estimation Download PDF

Info

Publication number
CN116843383A
CN116843383A CN202311120622.3A CN202311120622A CN116843383A CN 116843383 A CN116843383 A CN 116843383A CN 202311120622 A CN202311120622 A CN 202311120622A CN 116843383 A CN116843383 A CN 116843383A
Authority
CN
China
Prior art keywords
parameter
classification
excitation
user
loss function
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
Application number
CN202311120622.3A
Other languages
Chinese (zh)
Inventor
刘越
郑淳元
李昊轩
崔鹏
吴鹏
况琨
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Renmin University of China
Original Assignee
Renmin University of China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Renmin University of China filed Critical Renmin University of China
Priority to CN202311120622.3A priority Critical patent/CN116843383A/en
Publication of CN116843383A publication Critical patent/CN116843383A/en
Pending legal-status Critical Current

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
    • G06Q30/0207—Discounts or incentives, e.g. coupons or rebates
    • G06Q30/0239—Online discounts or incentives
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/27—Regression, e.g. linear or logistic regression
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0499—Feedforward networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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
    • G06Q30/0207—Discounts or incentives, e.g. coupons or rebates
    • G06Q30/0224—Discounts or incentives, e.g. coupons or rebates based on user history

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Strategic Management (AREA)
  • Finance (AREA)
  • Development Economics (AREA)
  • Accounting & Taxation (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Economics (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Probability & Statistics with Applications (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a personalized incentive method and a device based on inverse fact identification and estimation, wherein the method comprises the following steps: training a preset first model and a preset second model based on a preset training data set, wherein the first model outputs probability parameters of excitation; respectively inputting the user object pairs to be matched into a second model, and outputting classification parameters of the users by the second model; the classification parameters comprise a first classification parameter and a second classification parameter, the first class parameter and the second class parameter are calculated based on the first classification parameter and the second classification parameter, and the excitation coefficient is calculated based on the first class parameter, the second class parameter, a first excitation coefficient preset corresponding to the first class parameter and a second excitation coefficient preset corresponding to the second class parameter; a determination is made as to whether to issue an incentive for the user of the user item pair based on the incentive coefficient.

Description

Individualized excitation method and device based on counterfactual identification and estimation
Technical Field
The invention relates to the technical field of computers, in particular to a personalized incentive method and device based on inverse fact identification and estimation.
Background
The conversion feedback directly reflects the user preferences and is associated with the total merchandise transaction amount. To attract user interest and increase platform revenue, many e-commerce companies provide users with personalized incentive measures (e.g., sending coupons or giving cash returns) to increase conversion rates. The effective incentive measure for the specific consumer can increase the viscosity of the user, realize the increase of the user, and further realize the win-win situation of the user and the e-commerce company. Therefore, motivational measures are widely adopted in many application scenarios such as e-commerce transactions and music websites.
Typically, a personalized incentive strategy will provide incentives for specific sub-populations based on observed user and item characteristics. Some users will accept the incentive and some users will ignore it, so for users who ignore the incentive, the incentive is easy to reduce the user experience, and therefore, the incentive needs to be pushed accurately to improve the user experience.
Disclosure of Invention
In view of this, embodiments of the present invention provide a personalized incentive method based on anti-facts identification and estimation to obviate or ameliorate one or more of the disadvantages of the prior art.
One aspect of the present invention provides a personalized incentive method based on anti-facts identification and estimation, the method comprising the steps of:
Training a first model and a second model which are preset based on a training data set, constructing a loss function based on output values of the first model and the second model in the training process, and training the first model and the second model, wherein the first model is based on probability parameters of input user articles on output user on output excitation;
respectively inputting the user object pairs to be matched into a second model, wherein the second model outputs classification parameters of the user based on the user object pairs;
the classification parameters comprise a first classification parameter and a second classification parameter, the first classification parameter and the second classification parameter are calculated based on the first classification parameter and the second classification parameter, the excitation coefficient is calculated based on the first classification parameter, the second classification parameter, a first excitation coefficient preset corresponding to the first classification parameter and a second excitation coefficient preset corresponding to the second classification parameter, the first classification parameter corresponds to the classification parameter of the user exciting the purchaser, and the second classification parameter corresponds to the label and is always the purchaser;
a determination is made as to whether to issue an incentive for the user of the user item pair based on the incentive coefficient.
By adopting the scheme, the category parameters of the user object pair are judged based on the first category parameters and the second category parameters, but the category parameters are often output in various category categories for the same user model in the actual process, the user category is judged simply by the size of each category parameter, and whether the user is stimulated or not is judged by the user category, so that the user which does not want to accept the stimulus is received, the user experience is poor, the excitation coefficient is calculated by combining the first category parameters and the second category parameters, and whether the user of the user object pair is stimulated or not is judged based on the excitation coefficient, so that the excitation condition can be accurately given to the user which needs, and the user experience is improved.
In some embodiments of the present invention, in the step of determining whether to issue an incentive to the user of the user item pair based on the incentive coefficient, it is determined that the incentive is issued to the user of the user item pair if the incentive coefficient is a positive number, and it is determined that the incentive is not issued to the user of the user item pair if the incentive coefficient is not a positive number.
In some embodiments of the present invention, in the step of calculating the excitation coefficient based on the first class parameter, the second class parameter, the first excitation coefficient preset corresponding to the first class parameter, and the second excitation coefficient preset corresponding to the second class parameter, the excitation coefficient is calculated according to the following formula:
wherein ,the excitation coefficients are represented, d represents the first class parameter, e represents the second class parameter, s1 represents the first excitation coefficient preset corresponding to the first class parameter, and s2 represents the second excitation coefficient preset corresponding to the second class parameter.
By adopting the scheme, the label corresponding to the second category parameter is an always-buyers, the always-buyers do not need to be excited, and the user experience is easily reduced if the always-buyers are excited, so that the product of the second category parameter and the second excitation coefficient preset corresponding to the second category parameter is used as negative excitation, the excitation coefficient is calculated by combining the first category parameter and the second category parameter, and the user experience is improved.
In some embodiments of the present invention, the classification parameters further include a third classification parameter, where the third classification parameter is a third classification parameter, and the third classification parameter corresponds to a classification parameter of a user of the excitation recipient, and in the step of calculating the excitation coefficient based on the first classification parameter, the second classification parameter, a first excitation coefficient preset corresponding to the first classification parameter, and a second excitation coefficient preset corresponding to the second classification parameter, the excitation coefficient is further calculated based on the third classification parameter and the third excitation coefficient preset corresponding to the third classification parameter.
In some embodiments of the present invention, in the step of calculating the excitation coefficients further based on the third classification parameter and the third excitation coefficients preset corresponding to the third classification parameter, the excitation coefficients are calculated according to the following formula:
wherein ,represents an excitation coefficient, d represents a first class parameter, e represents a second class parameter, s1 represents a first excitation coefficient preset corresponding to the first class parameter, s2 represents a second excitation coefficient preset corresponding to the second category parameter, c represents a third category parameter, and s3 represents a third excitation coefficient preset corresponding to the third category parameter.
By adopting the scheme, the label corresponding to the third category parameter is the excitation receiver, the excitation receiver is a user who can accept the issued excitation, but can not purchase the object no matter whether the excitation is issued or not, so that the excitation receiver is not required to issue the excitation, and therefore, the product of the third category parameter and the third excitation coefficient preset corresponding to the third category parameter is used as negative excitation, and the excitation coefficient is calculated by combining the first category parameter, the second category parameter and the third category parameter, so that the user experience is further improved.
In some embodiments of the present invention, the first classification parameter is a label of a user who purchases an item whenever an incentive is issued, the second classification parameter is a label of a user who purchases an item without an incentive, and in the step of calculating the first classification parameter and the second classification parameter based on the first classification parameter and the second classification parameter, a difference between the first classification parameter and the second classification parameter is calculated as the first classification parameter.
In some embodiments of the present invention, the classification parameters further include a fourth classification parameter, where the fourth classification parameter is a fourth classification parameter, and the fourth classification parameter corresponds to a label of a user who is never a recipient, and in the step of calculating the first classification parameter and the second classification parameter based on the first classification parameter and the second classification parameter, a difference between the second classification parameter and the fourth classification parameter is calculated as the second classification parameter.
In some embodiments of the present invention, the classification parameters further include a fifth classification parameter, where the fifth classification parameter is a fifth classification parameter, training a preset first model and a preset second model based on a preset training data set, constructing a loss function based on output values of the first model and the second model during training, and calculating a first loss function based on the output value of the first model, the fifth classification parameter and the first label value during training the first model and the second model; calculating a second loss function based on the output value of the first model, the fourth class parameter and the second label value; calculating a third loss function based on the output value of the first model, the third class parameter and the third label value; calculating a fourth loss function based on the output value of the first model, the second class parameter, the fourth class parameter, and the fourth label value; a fifth loss function is calculated based on the output value of the first model, the first class parameter, the third class parameter, the fifth class parameter, and the fifth label value, a total loss function is calculated based on the first loss function, the second loss function, the third loss function, the fourth loss function, and the fifth loss function, and the first model and the second model are trained based on the total loss function.
In some embodiments of the present invention, in the step of calculating the total loss function based on the first loss function, the second loss function, the third loss function, the fourth loss function, and the fifth loss function, the total loss function is calculated based on the following formula:
wherein ,representing the total loss function of the device,L1 denotes a first loss function,L2 denotes a second loss function,L3 denotes a third loss function,L4 denotes a fourth loss function,Land 5 denotes a fifth loss function.
In an implementation, the first loss function is expressed asThe second loss function is expressed as +.>The third loss function is expressed asThe fourth loss function is expressed asThe fifth loss function is expressed as，/>Representing a first tag value, a second tag value, a third tag value, a fourth tag value and a fifth tag value, respectively, +.>Output value representing the first model, +.>Representing a fifth category parameter, ">Representing a fourth category parameter, ">Representing a third category parameter, ">Representing a second category parameter, ">Representing a first class parameter.
In the course of the specific implementation process, the method comprises,representation->And->Constructing a loss function; />Representation->And->Constructing a loss function;representation->And->Constructing a loss function; Representation->And->Constructing a loss function; />Representation ofAnd->And constructing a loss function.
In a specific implementation process, the loss functions can be calculated by a cross entropy mode and the like, and the loss functions are calculated by the methodIs the representation of (i.e. calculate->A loss function therebetween.
The second aspect of the invention also provides a personalized incentive apparatus based on counterfactual identification and estimation, the apparatus comprising a computer device comprising a processor and a memory, the memory having stored therein computer instructions for executing the computer instructions stored in the memory, the apparatus implementing the steps of the method as hereinbefore described when the computer instructions are executed by the processor.
The third aspect of the present invention also provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps performed by the aforementioned anti-facts identification and estimation based personalized incentive method.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention may be realized and attained by the structure particularly pointed out in the written description and drawings.
It will be appreciated by those skilled in the art that the objects and advantages that can be achieved with the present application are not limited to the above-described specific ones, and that the above and other objects that can be achieved with the present application will be more clearly understood from the following detailed description.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate and together with the description serve to explain the application.
FIG. 1 is a schematic diagram of one embodiment of a personalized incentive method based on anti-facts identification and estimation of the present application;
FIG. 2 is a schematic diagram of the structure of the model of the present application.
Detailed Description
The present application will be described in further detail with reference to the following embodiments and the accompanying drawings, in order to make the objects, technical solutions and advantages of the present application more apparent. The exemplary embodiments of the present application and the descriptions thereof are used herein to explain the present application, but are not intended to limit the application.
It should be noted here that, in order to avoid obscuring the present application due to unnecessary details, only structures and/or processing steps closely related to the solution according to the present application are shown in the drawings, while other details not greatly related to the present application are omitted.
Introduction to the prior art:
the first prior art relates to a method for classifying users in different inputted user object pairs through a neural network model, and outputting classification parameters of the users for each class through a classification layer in the neural network model, when the classification parameters of the users in the class of coupon buyers are larger than a threshold value, judging the users as coupon buyers, and issuing incentives to the users. However, in the first prior art, though the coupon purchaser can be accurately identified, in the first prior art, the classification parameters of the user in other categories are ignored, a certain error often exists in the classification, the influence of other classification parameters on whether the incentive should be issued is ignored, and if the user actually belongs to an always-purchasers or an incentive receiver, the user receives unnecessary message pushing of the incentive, so that the user experience is easily reduced.
To solve the above problems, as shown in fig. 1, the present invention proposes a personalized incentive method based on anti-facts identification and estimation, the steps of the method include:
step S100, training a preset first model and a preset second model based on a preset training data set, constructing a loss function based on output values of the first model and the second model in the training process, and training the first model and the second model, wherein the first model is based on probability parameters of input user articles on output excitation of an output user;
In an implementation, the first model and the second model are both feedforward neural network models, which are also referred to as multi-layer neural networks (Multilayer Perceptron, MLP) in which information is propagated forward in the neural network.
In a specific implementation, the incentive may be a coupon or a cash reward, where the coupon is a coupon that can only be used by a purchase, and the cash reward is directly issued cash.
As shown in fig. 2, the tendency score prediction model in fig. 2 is the first model, the main hierarchical prediction model is the second model, the tendency score prediction model performs up-dimensional stitching on the characteristics of the user and the article, then inputs the result into a feedforward neural network (MLP) for training, and finally outputs probability parameters of the user on output excitation; the master hierarchical prediction model outputs probabilities for five people, "never purchaser", "never receiver", "coupon purchaser", and "never purchaser". Specifically, the main hierarchical prediction model performs up-dimensional stitching on the features of the user and the object, inputs the features into a feedforward neural network (MLP), and finally outputs the probability that the user belongs to five layers. The parameter sharing mechanism means that the tendency score prediction model and the main hierarchical prediction model share model parameters at the time of up-dimension stitching.
Step S200, respectively inputting the user object pairs to be matched into a second model, and outputting classification parameters of the user based on the user object pairs by the second model;
the article user pair is the combination of the article and the user, and the first model is based on outputting the combination of the article and the user to output excitation, namely probability parameters, to the user.
Step S300, the classification parameters comprise a first classification parameter and a second classification parameter, the first classification parameter and the second classification parameter are calculated based on the first classification parameter and the second classification parameter, the excitation coefficient is calculated based on the first classification parameter, the second classification parameter, a first excitation coefficient preset corresponding to the first classification parameter and a second excitation coefficient preset corresponding to the second classification parameter, the first classification parameter corresponds to the classification parameter of the user exciting the purchaser, and the second classification parameter corresponds to the label and is always the purchaser;
step S400, determining whether to issue an incentive for the user of the user item pair based on the incentive coefficient.
By adopting the scheme, the category parameters of the user object pair are judged based on the first category parameters and the second category parameters, but the category parameters are often output in various category categories for the same user model in the actual process, the user category is judged simply by the size of each category parameter, and whether the user is stimulated or not is judged by the user category, so that the user which does not want to accept the stimulus is received, the user experience is poor, the excitation coefficient is calculated by combining the first category parameters and the second category parameters, and whether the user of the user object pair is stimulated or not is judged based on the excitation coefficient, so that the excitation condition can be accurately given to the user which needs, and the user experience is improved.
In some embodiments of the present invention, in the step of determining whether to issue an incentive to the user of the user item pair based on the incentive coefficient, it is determined that the incentive is issued to the user of the user item pair if the incentive coefficient is a positive number, and it is determined that the incentive is not issued to the user of the user item pair if the incentive coefficient is not a positive number.
In some embodiments of the present invention, in the step of calculating the excitation coefficient based on the first class parameter, the second class parameter, the first excitation coefficient preset corresponding to the first class parameter, and the second excitation coefficient preset corresponding to the second class parameter, the excitation coefficient is calculated according to the following formula:
wherein ,represents the excitation coefficient, d represents the first class parameter, e represents the first class parameterThe second-class parameters, s1 represents a first excitation coefficient preset corresponding to the first-class parameters, and s2 represents a second excitation coefficient preset corresponding to the second-class parameters.
By adopting the scheme, the label corresponding to the second category parameter is an always-buyers, the always-buyers do not need to be excited, and the user experience is easily reduced if the always-buyers are excited, so that the product of the second category parameter and the second excitation coefficient preset corresponding to the second category parameter is used as negative excitation, the excitation coefficient is calculated by combining the first category parameter and the second category parameter, and the user experience is improved.
In some embodiments of the present invention, the classification parameters further include a third classification parameter, where the third classification parameter is a third classification parameter, and the third classification parameter corresponds to a classification parameter of a user of the excitation recipient, and in the step of calculating the excitation coefficient based on the first classification parameter, the second classification parameter, a first excitation coefficient preset corresponding to the first classification parameter, and a second excitation coefficient preset corresponding to the second classification parameter, the excitation coefficient is further calculated based on the third classification parameter and the third excitation coefficient preset corresponding to the third classification parameter.
In an implementation, the incentive receiver receives an incentive if the incentive is issued but does not purchase the commodity, and does not purchase the class of the user of the commodity if the incentive is not issued.
In some embodiments of the present invention, in the step of calculating the excitation coefficients further based on the third classification parameter and the third excitation coefficients preset corresponding to the third classification parameter, the excitation coefficients are calculated according to the following formula:
wherein ,represents an excitation coefficient, d represents a first class parameter, e represents a second class parameter, s1 represents a first excitation coefficient preset corresponding to the first class parameter, and s2 represents And c represents a third class parameter, and s3 represents a third excitation coefficient preset corresponding to the third class parameter.
By adopting the scheme, the label corresponding to the third category parameter is the excitation receiver, the excitation receiver is a user who can accept the issued excitation, but can not purchase the object no matter whether the excitation is issued or not, so that the excitation receiver is not required to issue the excitation, and therefore, the product of the third category parameter and the third excitation coefficient preset corresponding to the third category parameter is used as negative excitation, and the excitation coefficient is calculated by combining the first category parameter, the second category parameter and the third category parameter, so that the user experience is further improved.
In some embodiments of the present invention, the first classification parameter is a label of a user who purchases an item whenever an incentive is issued, the second classification parameter is a label of a user who purchases an item without an incentive, and in the step of calculating the first classification parameter and the second classification parameter based on the first classification parameter and the second classification parameter, a difference between the first classification parameter and the second classification parameter is calculated as the first classification parameter.
In the specific implementation process, the first category parameter corresponds to the category parameter of the user of the incentive purchaser, namely, the class of the user who gets the incentive and purchases the commodity if the incentive purchaser is issued with the incentive; the label corresponding to the second category parameter is an always-buying party, the always-buying party obtains the incentive and purchases the commodity if the incentive is issued, and the category of the user who purchases the commodity if the incentive is not issued.
By adopting the scheme, the excitation purchaser cannot directly classify and calculate by using the label, and the scheme is convenient for calculating the corresponding parameter of the excitation purchaser by calculating the difference between the first classification parameter and the second classification parameter as the first classification parameter, namely, the first classification parameter=the first classification parameter-the second classification parameter.
In some embodiments of the present invention, the classification parameters further include a fourth classification parameter, where the fourth classification parameter is a fourth classification parameter, and the fourth classification parameter corresponds to a label of a user who is never a recipient, and in the step of calculating the first classification parameter and the second classification parameter based on the first classification parameter and the second classification parameter, a difference between the second classification parameter and the fourth classification parameter is calculated as the second classification parameter.
In an implementation, the never receiver does not receive the incentive if it is issued, but purchases the commodity, and also purchases the class of the commodity if it is not issued.
By adopting the scheme, the forever purchasers cannot directly classify and calculate the labels, and the method is convenient for calculating the corresponding parameters of the forever purchasers by calculating the difference between the second classification parameter and the fourth classification parameter as the second classification parameter, namely, the second classification parameter=the second classification parameter-the fourth classification parameter.
In some embodiments of the present invention, the classification parameters further include a fifth classification parameter, where the fifth classification parameter is a fifth classification parameter, training a preset first model and a preset second model based on a preset training data set, constructing a loss function based on output values of the first model and the second model during training, and calculating a first loss function based on the output value of the first model, the fifth classification parameter and the first label value during training the first model and the second model; calculating a second loss function based on the output value of the first model, the fourth class parameter and the second label value; calculating a third loss function based on the output value of the first model, the third class parameter and the third label value; calculating a fourth loss function based on the output value of the first model, the second class parameter, the fourth class parameter, and the fourth label value; a fifth loss function is calculated based on the output value of the first model, the first class parameter, the third class parameter, the fifth class parameter, and the fifth label value, a total loss function is calculated based on the first loss function, the second loss function, the third loss function, the fourth loss function, and the fifth loss function, and the first model and the second model are trained based on the total loss function.
In the implementation process, the label corresponding to the fifth category parameter is a label of a user who never buys the product, and the never buys the product even if the user is excited, the user does not get the excitation and does not purchase the product even if the user is excited, and the category of the user who does not purchase the product even if the user is not excited.
In a specific implementation process, the excitation coefficients corresponding to the first category parameter, the second category parameter, the third category parameter, the fourth category parameter and the fifth category parameter may be as shown in the following table one, which shows a middle levelCorresponding to the fifth category parameter,>corresponding to the fourth category parameter,>corresponding to the third category parameter->Corresponding to the first category parameter->Corresponding to the second category parameter, C (0) indicates whether the non-emitted excitation received the excitation, C (0) =0 indicates that the non-emitted excitation did not receive the excitation; c (1) indicates whether the emitted excitation received the excitation, C (1) =0 indicates that the emitted excitation did not receive the excitation, and C (1) =1 indicates that the emitted excitation received the excitation; y (0) indicates whether the non-issued incentive purchased the commodity, Y (0) =0 indicates that the non-issued incentive purchased the commodity, Y (0) =1 indicates that the non-issued incentive purchased the commodity, Y (1) indicates whether the issued incentive purchased the commodity, Y (1) =0 indicates that the issued incentive purchased the commodity, and Y (1) =1 indicates that the issued incentive purchased the commodity.
The present invention divides the user-item pairs into five tiers from a counterfactual perspective, referred to as "never purchasers", "never acceptors", "coupon purchasers" and "never purchasers", respectively, i.e., based on the combined potential results of the same individual.
List one
In some embodiments of the present invention, when the incentive of the application scenario is cash incentive, setting an incentive coefficient corresponding to the third category parameter to s3, and calculating the incentive coefficient according to the following formula:
wherein ,represents an excitation coefficient, d represents a first class parameter, e represents a second class parameter, s1 represents a first excitation coefficient preset corresponding to the first class parameter, s2 represents a second excitation coefficient preset corresponding to the second category parameter, c represents a third category parameter, and s3 represents a third excitation coefficient preset corresponding to the third category parameter.
In a specific implementation, the first excitation coefficient, the second excitation coefficient, and the third excitation coefficient may all be adjusted based on the characteristics of the user.
By adopting the scheme, the product of the third category parameter and the third excitation coefficient preset corresponding to the third category parameter is used as negative excitation, the excitation coefficient is calculated by combining the first category parameter, the second category parameter and the third category parameter, the user experience can be further improved, when the excitation of the application scene is a coupon, the coupon acceptor can directly obtain the reward, the loss is caused for the merchant, the benefit of the merchant can be ensured by applying the formula, and the win-win of the merchant and the user can be ensured.
In some embodiments of the present invention, when the excitation of the application scenario is a coupon, the excitation coefficient corresponding to the third category parameter is set to 0, and the excitation coefficient is calculated according to the following formula:
wherein ,the excitation coefficients are represented, d represents the first class parameter, e represents the second class parameter, s1 represents the first excitation coefficient preset corresponding to the first class parameter, and s2 represents the second excitation coefficient preset corresponding to the second class parameter.
By adopting the scheme, firstly, the product of the second class parameter and the second excitation coefficient preset corresponding to the second class parameter is used as negative excitation, the excitation coefficient is calculated by combining the first class parameter and the second class parameter, the user experience is improved, and if the excitation is an application scene of the coupon, the user can only apply the excitation after purchasing the commodity, so that the loss is not caused to the merchant if the commodity is not purchased, and the excitation coefficient corresponding to the third class parameter is set to be 0 when the excitation of the application scene is the coupon.
In the implementation process, the effective personalized incentive can improve the user experience and increase the platform income, thereby realizing the win-win situation of the user and the electronic commerce company. The prior art estimates the conditional average treatment effect of the user on the stimuli by categorizing the user, and then allocates the stimuli by maximizing the estimated treatment effect sum under a limited budget. However, some users purchase whether or not incentives are given, but if incentives are provided they will be actively picked up and used, we call "always purchaser". Identifying and predicting these "always purchasers" and reducing the incentive provided to them may enable more reasonable distribution of incentives. The invention divides the users into five layers from the perspective of individual counterfactual, and overcomes the failure of the prior art in identifying and predicting the 'always purchaser'.
In some embodiments of the present invention, in the step of calculating the total loss function based on the first loss function, the second loss function, the third loss function, the fourth loss function, and the fifth loss function, the total loss function is calculated based on the following formula:
wherein ,representing the total loss function of the device,L1 denotes a first loss function,L2 denotes a second loss function,L3 denotes a third loss function,L4 denotes a fourth loss function,Land 5 denotes a fifth loss function.
In an implementation, the first loss function is expressed asThe second loss function is expressed as +.>The third loss function is expressed asThe fourth loss function is expressed asThe fifth loss function is expressed as，/>Representing a first tag value, a second tag value, a third tag value, a fourth tag value and a fifth tag value, respectively, +.>Output value representing the first model, +.>Representing a fifth category parameter, ">Representing a fourth category parameter, ">Representing a third category parameter, ">Representing a second category parameter, ">Representing a first class parameter.
In the course of the specific implementation process, the method comprises,representation->And->Constructing a loss function; />Representation->And->Constructing a loss function;representation->And->Constructing a loss function; Representation->And->Constructing a loss function; />Representation ofAnd->And constructing a loss function.
In a specific implementation process, model parameters of the first model and the second model are updated by adopting a back propagation method based on the obtained total loss function value.
In a specific implementation, the first model and the second model are co-trained. Compared with independent training of multiple regression and trend models to perform strategy learning, the scheme can relieve the problems of data sparsity and deviation amplification, so that more accurate hierarchical prediction is realized. Furthermore, we also give theoretical guarantees of the lower return limit of the learned personalized incentive strategy. We conducted extensive experiments on three real world datasets, including two common motivational scenarios. Experimental results show that the learning method provided by the invention can accurately realize the inverse fact prediction of the individual, thereby realizing more effective personalized incentive strategy in practice and increasing social benefit.
The beneficial effect of this scheme includes:
1. the invention redefines the personalized incentive strategy learning problem from the inverse facts, overcomes the limitation of the method that the user class is judged by the size of each classification parameter in the prior art, and further judges whether to send incentive to the user through the user class;
2. The invention further provides a full-space multitask learning method of the counterfactual, the first model and the second model can be trained at the same time, and personalized excitation strategy learning can be performed more accurately.
3. The method is suitable for two common personalized excitation situations and has effectiveness.
Experimental example:
the present invention has been tested on three real world datasets.
The specific experimental procedure and data set are presented below: three real world datasets include Yelp, ML-1M, and KuaiRec.
The yellow dataset had 25,677 users, 25,815 items, and 731,671 user-item interactions were collected.
The ML-1M dataset contains 6,040 users, 3,952 items, and 1,000,209 user-item interactions.
The KuaiRec dataset includes 1,411 users, 3,327 items, and 4,676,570 user-item interactions.
And further data tagging.
The experimental results are shown in table two, table three and table four below:
in the second, third and fourth tables, the first scheme is to output the probability of purchasing the commodity by the user through the preset neural network model to make a decision, and when the product of the probability of purchasing the commodity by the user and the profit is greater than the issuing cost of the incentive in the case of issuing the incentive, the incentive is issued, otherwise, the incentive is not issued.
And secondly, calculating the classification value of the user in each category by adopting a result regression (outcome regression, OR) method, and judging whether to issue incentives to the user OR not based on the classification value of the user in the category of the coupon purchaser.
And thirdly, calculating the classification value of the user in each category by adopting a result regression (outcome regression, OR) method, wherein the classification value corresponds to the category parameter of each category in the scheme, calculating the excitation coefficient by adopting the method of the scheme, judging that the user of the user object pair is excited if the excitation coefficient is positive, and judging that the user of the user object pair is not excited if the excitation coefficient is not positive.
And in the fourth scheme, a method of inverse tendency scoring (Inverse Propensity Scoring, IPS) is adopted to calculate the classification value of the user in each category, and whether to issue incentives to the user is judged based on the classification value of the user in the category of the coupon purchaser.
The fifth scheme is that the classification value of the user in each category is calculated by adopting a reverse tendency scoring (Inverse Propensity Scoring, IPS) method, the excitation coefficient is calculated by adopting the method of the scheme, if the excitation coefficient is positive, the excitation is judged to be issued to the user of the user article pair, and if the excitation coefficient is not positive, the excitation is judged not to be issued to the user of the user article pair.
The sixth scheme is to calculate the classification value of the user in each category by adopting a Double Robust (DR) method, and determine whether to issue incentives to the user based on the classification value of the user in the category of the coupon purchaser.
The seventh scheme is that a Double Robust (DR) method is adopted to calculate the classification value of the user in each category, the method of the present scheme is adopted to calculate the excitation coefficient, if the excitation coefficient is a positive number, the excitation is determined to be issued to the user of the user article pair, and if the excitation coefficient is not a positive number, the excitation is determined to be not issued to the user of the user article pair.
The eighth scheme is that the method comprises the steps of training a first model and a second model which are preset, constructing a loss function based on output values of the first model and the second model in the training process, and training the first model and the second model, wherein the first model is based on probability parameters of input user articles on output excitation of an output user; respectively inputting the user object pairs to be matched into a second model, wherein the second model outputs classification parameters of the user based on the user object pairs; the classification parameters comprise a first classification parameter and a second classification parameter, the first classification parameter and the second classification parameter are calculated based on the first classification parameter and the second classification parameter, the excitation coefficient is calculated based on the first classification parameter, the second classification parameter, a first excitation coefficient preset corresponding to the first classification parameter and a second excitation coefficient preset corresponding to the second classification parameter, the first classification parameter corresponds to the classification parameter of the user exciting the purchaser, and the second classification parameter corresponds to the label and is always the purchaser; if the excitation coefficient is positive, the excitation is judged to be issued to the user of the user object pair, and if the excitation coefficient is not positive, the excitation is judged not to be issued to the user of the user object pair.
In the coupon scenario, the excitation coefficients are calculated according to the following formula:
wherein ,representing excitation coefficients, d representing first class parameters, e representing second class parameters, s1 representing first excitation coefficients preset corresponding to the first class parameters, and s2 representing second excitation coefficients preset corresponding to the second class parameters; />
In a cash scenario, the excitation coefficients are calculated according to the following formula:
wherein ,represents an excitation coefficient, d represents a first class parameter, e represents a second class parameter, s1 represents a first excitation coefficient preset corresponding to the first class parameter, s2 represents a second excitation coefficient preset corresponding to the second category parameter, c represents a third category parameter, and s3 represents a third excitation coefficient preset corresponding to the third category parameter.
In the coupon scenes of the second table, the third table and the fourth table, neutrality represents the sum of the number of people belonging to never purchasers, never acceptors and incentives acceptors in the users who issue incentives by the scheme; forward direction indicates the number of people belonging to the motivating purchaser among the users who issue motivations for the scheme; the negative direction indicates the number of people belonging to the permanent purchasers among the users who send out the motivations for the proposal;
in the cash scenes of the second, third and fourth tables, neutrality represents that the users of the scheme issuing incentives belong to never purchasers and never acceptors; forward direction indicates the number of people belonging to the motivating purchaser among the users who issue motivations for the scheme; the negative direction indicates the number of people belonging to the incentive recipients and the always-buyers among the users who issued the incentive for the scheme.
The return in the second, third and fourth tables is the value obtained by positive people count the profit of each positive person and negative people count the loss of each negative person;
in the implementation process, the profit of each positive person is set to be 1, and the loss of each negative person is set to be 0.4.
The rate of increase in Table II, table III and Table IV at scheme III represents the rate of increase of the rate of return of scheme III relative to the rate of increase of scheme IV at scheme V, the rate of increase in scheme seven represents the rate of increase of the rate of return of scheme seven relative to the rate of increase of scheme six, and the rate of increase in scheme eight represents the rate of increase of the rate of return of scheme eight relative to the maximum return of schemes III, five and seven.
Wherein, the second table is the experimental result based on the Yelp dataset, the third table is the experimental result based on the ML-1M dataset, and the fourth table is the experimental result based on the KuaiRec dataset.
Watch II
Watch III
Table four
From the above, it can be seen from tables two, three and four that scheme one performs worst under all scenarios and data sets, while schemes two, four and six perform slightly higher than scheme one. This demonstrates the importance of estimating the effect of personalized incentives. Next, all the proposed inverse estimation methods, schemes three, five and seven, are significantly improved over schemes two, four and six. This is because the proposed anti-facts estimation method can identify and estimate the probability that an individual belongs to each anti-facts category, while schemes two, four and six can identify and estimate the probability that an individual belongs to the "coupon purchaser" category. The proposed inverse-fact multitasking learning method, i.e. the present solution, then exhibits the best performance under all scenarios and data sets. Notably, on Yelp and ML-1M, the overall excitation improvement of the method exceeds 50% of the optimal values in schemes three, five and seven, thereby enabling more accurate and robust predictions.
The embodiment of the invention also provides a personalized incentive device based on the counterfactual identification and estimation, which comprises a computer device, wherein the computer device comprises a processor and a memory, the memory is stored with computer instructions, the processor is used for executing the computer instructions stored in the memory, and the device realizes the steps realized by the method when the computer instructions are executed by the processor.
The embodiments of the present invention also provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps performed by the aforementioned anti-facts identification and estimation based personalized incentive method. The computer readable storage medium may be a tangible storage medium such as Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, floppy disks, hard disk, a removable memory disk, a CD-ROM, or any other form of storage medium known in the art.
Those of ordinary skill in the art will appreciate that the various illustrative components, systems, and methods described in connection with the embodiments disclosed herein can be implemented as hardware, software, or a combination of both. The particular implementation is hardware or software dependent on the specific application of the solution and the design constraints. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, a plug-in, a function card, or the like. When implemented in software, the elements of the invention are the programs or code segments used to perform the required tasks. The program or code segments may be stored in a machine readable medium or transmitted over transmission media or communication links by a data signal carried in a carrier wave.
It should be understood that the invention is not limited to the particular arrangements and instrumentality described above and shown in the drawings. For the sake of brevity, a detailed description of known methods is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method processes of the present invention are not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between steps, after appreciating the spirit of the present invention.
In this disclosure, features that are described and/or illustrated with respect to one embodiment may be used in the same way or in a similar way in one or more other embodiments and/or in combination with or instead of the features of the other embodiments.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, and various modifications and variations can be made to the embodiments of the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A personalized incentive method based on counterfactual identification and estimation, characterized in that the method comprises the steps of:
training a first model and a second model which are preset based on a training data set, constructing a loss function based on output values of the first model and the second model in the training process of the second model, and training the first model and the second model, wherein the first model is based on probability parameters of input user articles on output excitation of an output user;
respectively inputting the user object pairs to be matched into a second model, wherein the second model outputs classification parameters of the user based on the user object pairs;
the classification parameters comprise a first classification parameter and a second classification parameter, the first classification parameter and the second classification parameter are calculated based on the first classification parameter and the second classification parameter, the excitation coefficient is calculated based on the first classification parameter, the second classification parameter, a first excitation coefficient preset corresponding to the first classification parameter and a second excitation coefficient preset corresponding to the second classification parameter, the first classification parameter corresponds to the classification parameter of the user exciting the purchaser, and the second classification parameter corresponds to the label and is always the purchaser;
A determination is made as to whether to issue an incentive for the user of the user item pair based on the incentive coefficient.
2. The method for personalized incentive based on anti-facts identification and estimation according to claim 1, wherein in the step of deciding whether to issue incentive for the user of the user item pair based on the incentive coefficient, if the incentive coefficient is a positive number, it is decided that incentive is issued for the user of the user item pair, and if the incentive coefficient is not a positive number, it is decided that incentive is not issued for the user of the user item pair.
3. The method according to claim 2, wherein in the step of calculating the excitation coefficients based on the first class parameter, the second class parameter, the first excitation coefficients preset corresponding to the first class parameter, and the second excitation coefficients preset corresponding to the second class parameter, the excitation coefficients are calculated according to the following formula:
wherein ,represents an excitation coefficient, d represents a first class parameter, e represents a second class parameter, s1 represents a first excitation coefficient preset corresponding to the first class parameter, and s2 represents a second excitation coefficient preset corresponding to the second class parameter。
4. The method according to claim 2, wherein the classification parameters further include a third classification parameter, the third classification parameter being a third classification parameter, the third classification parameter corresponding to a classification parameter of a user of the excitation recipient, wherein in the step of calculating the excitation coefficients based on the first classification parameter, the second classification parameter, a first excitation coefficient preset corresponding to the first classification parameter, and a second excitation coefficient preset corresponding to the second classification parameter, the excitation coefficients are further calculated based on the third classification parameter and a third excitation coefficient preset corresponding to the third classification parameter.
5. The method for personalized excitation based on inverse fact identification and estimation according to claim 4, wherein in the step of calculating the excitation coefficients further based on the third classification parameters and the third excitation coefficients preset corresponding to the third classification parameters, the excitation coefficients are calculated according to the following formula:
wherein ,represents an excitation coefficient, d represents a first class parameter, e represents a second class parameter, s1 represents a first excitation coefficient preset corresponding to the first class parameter, s2 represents a second excitation coefficient preset corresponding to the second category parameter, c represents a third category parameter, and s3 represents a third excitation coefficient preset corresponding to the third category parameter.
6. The method of claim 1 to 5, wherein the first classification parameter is a label of a user who purchases an item whenever an incentive is issued, the second classification parameter is a label of a user who is not issued an incentive but purchases an item, and the difference between the first classification parameter and the second classification parameter is calculated as the first classification parameter in the step of calculating the first classification parameter and the second classification parameter based on the first classification parameter and the second classification parameter.
7. The method according to claim 4, wherein the classification parameters further include a fourth classification parameter, the fourth classification parameter being a fourth classification parameter, the fourth classification parameter corresponding to a label of a user who is never a recipient, and wherein in the step of calculating the first classification parameter and the second classification parameter based on the first classification parameter and the second classification parameter, a difference between the second classification parameter and the fourth classification parameter is calculated as the second classification parameter.
8. The method according to claim 7, wherein the classification parameters further include a fifth classification parameter, the fifth classification parameter is a fifth classification parameter, the training is performed on the first model and the second model based on the preset training data set, the loss function is constructed based on the output values of the first model and the second model during the training, and the first loss function is calculated based on the output value of the first model, the fifth classification parameter and the first label value during the training; calculating a second loss function based on the output value of the first model, the fourth class parameter and the second label value; calculating a third loss function based on the output value of the first model, the third class parameter and the third label value; calculating a fourth loss function based on the output value of the first model, the second class parameter, the fourth class parameter, and the fourth label value; a fifth loss function is calculated based on the output value of the first model, the first class parameter, the third class parameter, the fifth class parameter, and the fifth label value, a total loss function is calculated based on the first loss function, the second loss function, the third loss function, the fourth loss function, and the fifth loss function, and the first model and the second model are trained based on the total loss function.
9. The method of claim 8, wherein in the step of calculating the total loss function based on the first loss function, the second loss function, the third loss function, the fourth loss function, and the fifth loss function, the total loss function is calculated based on the following formula:
wherein ,representing the total loss function of the device,L1 denotes a first loss function,L2 denotes a second loss function,L3 denotes a third loss function,L4 denotes a fourth loss function,Land 5 denotes a fifth loss function.
10. A personalized incentive device based on counterfactual identification and estimation, characterized in that it comprises a computer device comprising a processor and a memory, said memory having stored therein computer instructions for executing the computer instructions stored in said memory, which device, when executed by the processor, implements the steps implemented by the method according to any of claims 1-9.
CN202311120622.3A 2023-09-01 2023-09-01 Individualized excitation method and device based on counterfactual identification and estimation Pending CN116843383A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311120622.3A CN116843383A (en) 2023-09-01 2023-09-01 Individualized excitation method and device based on counterfactual identification and estimation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311120622.3A CN116843383A (en) 2023-09-01 2023-09-01 Individualized excitation method and device based on counterfactual identification and estimation

Publications (1)

Publication Number Publication Date
CN116843383A true CN116843383A (en) 2023-10-03

Family

ID=88162057

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311120622.3A Pending CN116843383A (en) 2023-09-01 2023-09-01 Individualized excitation method and device based on counterfactual identification and estimation

Country Status (1)

Country Link
CN (1) CN116843383A (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060229946A1 (en) * 1996-12-23 2006-10-12 Scroggie Michael C System and method for providing shopping aids and incentives to customers through a computer network
CN105354728A (en) * 2015-12-11 2016-02-24 北京京东尚科信息技术有限公司 Discount coupon pushing method and device
CN108985809A (en) * 2017-06-02 2018-12-11 北京京东尚科信息技术有限公司 Motivate method, apparatus, electronic equipment and the storage medium of push
CN110245980A (en) * 2019-05-29 2019-09-17 阿里巴巴集团控股有限公司 The method and apparatus for determining target user's exiting form based on neural network model
CN110838019A (en) * 2018-08-17 2020-02-25 北京京东尚科信息技术有限公司 Method and device for determining trial supply distribution crowd

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060229946A1 (en) * 1996-12-23 2006-10-12 Scroggie Michael C System and method for providing shopping aids and incentives to customers through a computer network
CN105354728A (en) * 2015-12-11 2016-02-24 北京京东尚科信息技术有限公司 Discount coupon pushing method and device
CN108985809A (en) * 2017-06-02 2018-12-11 北京京东尚科信息技术有限公司 Motivate method, apparatus, electronic equipment and the storage medium of push
CN110838019A (en) * 2018-08-17 2020-02-25 北京京东尚科信息技术有限公司 Method and device for determining trial supply distribution crowd
CN110245980A (en) * 2019-05-29 2019-09-17 阿里巴巴集团控股有限公司 The method and apparatus for determining target user's exiting form based on neural network model

Similar Documents

Publication Publication Date Title
US11170395B2 (en) Digital banking platform and architecture
US20210035152A1 (en) Predicting the effectiveness of a marketing campaign prior to deployment
US10628837B2 (en) System and method for modeling customer behavior
US8620738B2 (en) Loyalty program incentive determination
US10037535B2 (en) Loyalty program parameter collaboration
US10430803B2 (en) Methods and systems for predicting consumer behavior from transaction card purchases
US20180046925A1 (en) Method, apparatus, and computer program product for determining closing metrics
US20110054981A1 (en) Analyzing Local Non-Transactional Data with Transactional Data in Predictive Models
US20090012839A1 (en) Determining Brand Affiliations
US20120271744A1 (en) Industry size of wallet
US20100145786A1 (en) Loyalty program service
US12014389B2 (en) Systems and methods for collaborative offer generation
WO2015079460A1 (en) System for computing contribution and providing appropriate incentives
US11328302B2 (en) Systems and methods using a data structure summarizing item information in authorization request messages for communication in transactions involving multiple items
US20160005072A1 (en) Method and system for online commerce analysis
US11205176B2 (en) Methods and systems for authorizing program activities
Hasumoto et al. Predicting customer churn for platform businesses: using latent variables of variational autoencoder as consumers’ purchasing behavior
US20230419184A1 (en) Causal Inference Machine Learning with Statistical Background Subtraction
Kazmi et al. Generic framework to predict repeat behavior of customers using their transaction history
AU2018306317A1 (en) System and method for detecting and responding to transaction patterns
CN116843383A (en) Individualized excitation method and device based on counterfactual identification and estimation
US11610222B1 (en) Lead user quality score
JP7307702B2 (en) Apparatus, program and method for estimating customer personality from customer information
US11763330B1 (en) System and method of recommending a payment option
US20230297826A1 (en) System and method for predicting transactional behavior in a network

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