CN113450146B - Multi-contact advertisement attribution method, system, computer device, and readable storage medium - Google Patents

Multi-contact advertisement attribution method, system, computer device, and readable storage medium Download PDF

Info

Publication number
CN113450146B
CN113450146B CN202110713526.4A CN202110713526A CN113450146B CN 113450146 B CN113450146 B CN 113450146B CN 202110713526 A CN202110713526 A CN 202110713526A CN 113450146 B CN113450146 B CN 113450146B
Authority
CN
China
Prior art keywords
crowd
attribution
model
contact
advertisement
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.)
Active
Application number
CN202110713526.4A
Other languages
Chinese (zh)
Other versions
CN113450146A (en
Inventor
王同乐
周星杰
李霞
孙泽懿
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.)
Shanghai Minglue Artificial Intelligence Group Co Ltd
Original Assignee
Shanghai Minglue Artificial Intelligence Group Co Ltd
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 Shanghai Minglue Artificial Intelligence Group Co Ltd filed Critical Shanghai Minglue Artificial Intelligence Group Co Ltd
Priority to CN202110713526.4A priority Critical patent/CN113450146B/en
Publication of CN113450146A publication Critical patent/CN113450146A/en
Application granted granted Critical
Publication of CN113450146B publication Critical patent/CN113450146B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06Q30/0241—Advertisements
    • G06Q30/0242—Determining effectiveness of advertisements
    • G06Q30/0245—Surveys
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Engineering & Computer Science (AREA)
  • Accounting & Taxation (AREA)
  • Development Economics (AREA)
  • Finance (AREA)
  • Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Information Transfer Between Computers (AREA)

Abstract

The present application relates to a multi-contact advertisement attribution method, system, computer device and computer readable storage medium, wherein the method comprises: a crowd-pack data set obtaining step, which is used for obtaining a data set comprising K crowd packs, wherein each crowd pack corresponds to the conversion rate of a crowd pack, and each crowd pack comprises J advertisement points; a data feature extraction step, namely carrying out feature extraction on each advertisement point in the data set to obtain a feature vector of each point; a model construction step, namely constructing a semi-black box multi-contact attribution model based on crowd-sourced data to represent the relation between each point position and crowd-sourced conversion rate, and determining model parameters based on the feature vectors; and an advertisement attribution step, wherein the prediction contribution of each advertisement point in the data set is calculated based on the semi-black box multi-contact attribution model, so that advertisement attribution is realized. The advertisement attribution is realized and the data privacy protection is ensured, so that the method and the device have higher feasibility.

Description

Multi-contact advertisement attribution method, system, computer device, and readable storage medium
Technical Field
The present application relates to the field of internet technology, and in particular, to a multi-contact advertisement attribution method, system, computer device, and computer readable storage medium.
Background
In recent years, online advertising has replaced traditional offline advertising, becoming the primary form of advertising. An advantage of online advertising is that advertisers will be able to obtain a large amount of user feedback and infer therefrom the impact of each ad spot on the overall ad campaign conversion population, a process of reverse inference of ad spot on ad campaign conversion conditions known as ad attribution. Advertisement attribution typically requires reconstruction of each user's browsing and clicking sequence, and then evaluation of the contribution of each advertisement spot based on the reconstructed sequence, which is beneficial to guiding advertisers to follow-up advertisement placement campaigns.
There are three commonly used advertising attribution algorithms. The first is an a/B test-based method, the second is a time series-based algorithm, and the third is a mixed model-based attribution method. A common disadvantage of these three types of methods is that they must use point location data specific to the user level, which is not workable for crowd-sourced data.
However, in recent years, due to factors such as laws and regulations in terms of data privacy protection and monopoly of the platform, it is difficult for a third party advertisement detection company to acquire advertisement data specific to the user level from an advertisement delivery platform (such as tremble, fast hands, etc.). The advertisement platform can only provide information such as the exposure quantity, the click quantity and the like of each advertisement in a crowd-sourced mode, and can not provide advertisement data specific to a user level.
As such, implementation of existing advertisement attribution based algorithms cannot be supported.
Disclosure of Invention
The embodiment of the application provides a multi-contact advertisement attribution method, a system, computer equipment and a computer readable storage medium, which are used for realizing the conversion contribution solving problem of point positions by collecting a plurality of groups of data based on crowd package level and training a semi-black box multi-contact attribution model, so that advertisement attribution is realized, and the data privacy protection is ensured while the advertisement attribution is realized, so that the application has higher feasibility.
In a first aspect, embodiments of the present application provide a multi-contact advertisement attribution method, including:
a crowd-pack data set obtaining step for obtaining a data set including K crowd-packs, each of the crowd-packs corresponding to a conversion rate of a crowd-packEach crowd-sourced includes J ad spots. Specifically, each advertisement spot comprises a plurality of exposure records, and each exposure record comprises spot information and the number of exposure released on the spot.
A data feature extraction step, wherein feature extraction is carried out on each advertisement point in the data set to obtain feature vectors of each pointSpecifically, the point location information of each point location is subjected to feature coding based on the One-Hot coding method.
A model construction step of constructing a semi-black box multi-contact attribution model based on crowd-sourced data to represent the relationship between each point location and crowd-sourced conversion rate, and based on the feature vectorDetermining model parameters;
an advertisement attribution step, namely calculating the prediction contribution of each advertisement point in the data set based on the semi-black box multi-contact attribution model to realize advertisement attribution;
wherein K, J is a natural number, K is {1,2, K, J e {1, 2.
Based on the steps, the advertisement attribution method can be based on coarse-granularity level data, particularly crowd-sourced level data, advertisement attribution can be realized without specific point location data at a user level, and the business environment of the current higher data privacy protection requirement is met.
In some of these embodiments, the method further comprises:
and a model evaluation step of acquiring the real contribution of each point bit based on the data set and evaluating the semi-black box multi-contact attribution model according to the real contribution and the predicted contribution, thereby evaluating the accuracy of the multi-contact advertising attribution method. Alternatively, R is used 2 The index evaluates the semi-black box multi-contact attribution model.
In some of these embodiments, the model building step further comprises:
a relation model construction step, namely constructing a half-black box multi-contact attribution model for representing the relation between each point in the crowd-sourced and the crowd-sourced conversion rate, wherein the half-black box multi-contact attribution model is as follows:
wherein,estimated conversion for the kth crowd-sourced,/->For the total exposure of the kth group of persons, n j Total number of exposure records for jth spot, +.>e (·) is by->Prediction b j W represents the network parameters of e (·), +.>The exposure quantity recorded for the ith exposure of the jth point of the k crowd pack;
a model parameter solving step of constructing a loss function and calculating the network parameters w and a parameters a by using a Gradient Descent method (Gradient Descent) on the loss function j 。
Based on the above, the embodiment of the application provides a semi-black box multi-contact attribution model, which realizes the prediction of the conversion rate of advertisement points by collecting a plurality of groups of data training based on crowd pack level, effectively utilizes discrete and continuous characteristics, and jointly realizes the conversion contribution solving problem of points by combining a neural network of a black box and a curve fitting method of a white box.
In some of these embodiments, the predicted contribution is calculated from the following model:
wherein,conversion exposure for the j-th spot, specifically: />
In some of these embodiments, the loss function is calculated from the following model:
wherein,estimated conversion for the kth crowd-sourced,/->Is the conversion rate of the kth crowd-sourced.
In a second aspect, embodiments of the present application provide a multi-contact advertisement attribution system for performing the multi-contact advertisement attribution method as described in the first aspect above, comprising:
the crowd-pack data set acquisition module is used for acquiring data sets comprising K crowd packs, wherein each crowd pack corresponds to the conversion rate of one crowd packEach crowd-sourced includes J ad spots. Specifically, each advertisement spot comprises a plurality of exposure records, and each exposure record comprises spot information and the number of exposure released on the spot.
The data feature extraction module is used for carrying out feature extraction on each advertisement point in the data set to obtain feature vectors of each pointSpecifically, the point location information of each point location is subjected to feature coding based on the One-Hot coding method.
The model construction module is used for constructing a semi-black box multi-contact attribution model based on crowd-sourced data so as to represent the relation between each point position and crowd-sourced conversion rate and based on the feature vectorDetermining model parameters;
the advertisement attribution module is used for calculating the prediction contribution of each advertisement point in the data set based on the semi-black box multi-contact attribution model to realize advertisement attribution;
wherein K, J is a natural number, K is {1,2, K, J e {1, 2.
Based on the module, the advertisement attribution method can be based on coarse-granularity level data, particularly crowd-sourced level data, advertisement attribution can be realized without specific point location data at a user level, and the business environment of the current higher data privacy protection requirement is met.
In some of these embodiments, the system further comprises:
and the model evaluation module is used for acquiring the real contribution of each bit based on the data set and evaluating the semi-black box multi-contact attribution model according to the real contribution and the predicted contribution so as to evaluate the accuracy of the multi-contact advertising attribution method. Alternatively, R is used 2 The index evaluates the semi-black box multi-contact attribution model.
In some of these embodiments, the model building module further comprises:
the relation model building module is used for building a half-black box multi-contact attribution model for representing the relation between each point in the crowd-sourced and the crowd-sourced conversion rate, and the half-black box multi-contact attribution model is as follows:
wherein,estimated conversion for the kth crowd-sourced,/->For the total exposure of the kth group of persons, n j Record total for exposure of jth spotCount (n)/(l)>e (·) is by->Prediction b j W represents the network parameters of e (·), +.>The exposure quantity recorded for the ith exposure of the jth point of the k crowd pack;
model parameter solving module, constructing a loss function and calculating the loss function by using gradient descent method to obtain the network parameters w and a parameters a j 。
Based on the above, the embodiment of the application provides a semi-black box multi-contact attribution model, which realizes the prediction of the conversion rate of advertisement points by collecting a plurality of groups of data training based on crowd pack level, effectively utilizes discrete and continuous characteristics, and jointly realizes the conversion contribution solving problem of points by combining a neural network of a black box and a curve fitting method of a white box.
In some of these embodiments, the predicted contribution is calculated from the following model:
wherein,conversion exposure for the j-th spot, specifically: />
In some of these embodiments, the loss function is calculated from the following model:
wherein,estimated conversion for the kth crowd-sourced,/->Is the conversion rate of the kth crowd-sourced.
In a third aspect, embodiments of the present application provide a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the multi-contact advertising attribution method as described in the first aspect above when executing the computer program.
In a fourth aspect, embodiments of the present application provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a multi-contact advertising attribution method as in the first aspect described above.
Compared with the related art, the multi-contact advertisement attribution method, the system, the computer equipment and the computer readable storage medium provided by the application especially relate to a marketing intelligent technology, and the semi-black box multi-contact attribution model based on crowd-sourced level data realizes advertisement attribution based on crowd-sourced level data, does not need to use user-sourced level data, has higher feasibility, fully utilizes self characteristics of advertisement points and exposure characteristics of the advertisement points, and achieves a very good prediction effect.
The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below to provide a more thorough understanding of the other features, objects, and advantages of the application.
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 application, illustrate embodiments of the application and together with the description serve to explain the application and do not constitute an undue limitation to the application. In the drawings:
FIG. 1 is a flow chart of a multi-contact advertising attribution method, according to an embodiment of the present application;
FIG. 2 is a block diagram of a multi-contact advertising attribution system, according to an embodiment of the present application;
FIG. 3 is another block diagram of a multi-contact advertising attribution system, according to an embodiment of the present application;
FIG. 4 is a flow chart of a multi-contact advertising attribution method, according to a preferred embodiment of the present application;
FIG. 5 is a graph of exposure versus conversion exposure defined in accordance with a preferred embodiment of the present application;
FIG. 6 is a semi-black box multi-contact attribution model organization structure diagram, according to a preferred embodiment of the present application;
FIG. 7 is a pseudo-code schematic diagram of a loss function solution according to a preferred embodiment of the present application.
Wherein:
1. a crowd-sourced data set acquisition module; 2. a data feature extraction module; 3. a model building module;
4. an advertisement attribution module; 5. a model evaluation module;
301. a relationship model construction module; 302. and a model parameter solving module.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described and illustrated below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden on the person of ordinary skill in the art based on the embodiments provided herein, are intended to be within the scope of the present application.
It is apparent that the drawings in the following description are only some examples or embodiments of the present application, and it is possible for those of ordinary skill in the art to apply the present application to other similar situations according to these drawings without inventive effort. Moreover, it should be appreciated that while such a development effort might be complex and lengthy, it would nevertheless be a routine undertaking of design, fabrication, or manufacture for those of ordinary skill having the benefit of this disclosure, and thus should not be construed as having the benefit of this disclosure.
Reference in the specification to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is to be expressly and implicitly understood by those of ordinary skill in the art that the embodiments described herein can be combined with other embodiments without conflict.
Unless defined otherwise, technical or scientific terms used herein should be given the ordinary meaning as understood by one of ordinary skill in the art to which this application belongs. Reference to "a," "an," "the," and similar terms herein do not denote a limitation of quantity, but rather denote the singular or plural. The terms "comprising," "including," "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, article, or apparatus that comprises a list of steps or modules (elements) is not limited to only those steps or elements but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. The terms "connected," "coupled," and the like in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The term "plurality" as used herein refers to two or more. "and/or" describes an association relationship of an association object, meaning that there may be three relationships, e.g., "a and/or B" may mean: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship. The terms "first," "second," "third," and the like, as used herein, are merely distinguishing between similar objects and not representing a particular ordering of objects.
The research on attribution problems mainly refers to the contribution rate of advertisements of all channels to user conversion, and the research on attribution problems can enable merchants to better optimize advertisement delivery strategies, reduce delivery cost and improve advertisement conversion rate. A good attribution model needs to be based on data, transparent enough to easily explain the principle of advertising in it. The user has an inherent interaction relationship between various advertisements contacted before conversion, so the single point attribution model must not solve the attribution problem well, and the attribution model established in consideration of all advertisement channels contacted by the user before the conversion is completed is called a Multi-contact attribution model (Multi-TouchAttribution, MTA).
The following explains the key terms involved in the current attribution of problems.
Conversion rate: conversion refers to the percentage of conversion exposure on the crowd sourcing (spot) to total exposure. The specific expression of the conversion rate is as follows:
wherein,representing the conversion exposure, ++>Indicating the total exposure.
Contribution: contribution means the conversion rate gp of each advertisement point to crowd-sourced group conv Percentage of contribution. The contribution specific expression is as follows:
wherein tp conv Indicating a point in a crowd-sourced, a transitionRate of conversion, gp conv Indicating the conversion rate of the crowd-sourced.
R 2 Value: also called a judgment coefficient, in linear regression, the ratio of the sum of squares of the interpretable dispersion to the sum of squares of the total dispersion is equal to the square of the correlation coefficient R, and is used to judge whether the regression model is good or bad, with the value range of [0,1 ]]。R 2 The specific expression of (2) is as follows:
wherein y is j The true value is represented by a value that is true,representing the predicted value.
In order to solve the problem that an existing advertisement attribution algorithm cannot work in a commercial environment with higher current user-level data privacy protection requirements, the application provides a semi-black box multi-contact attribution model based on crowd-sourced level data, and advertisement attribution based on the crowd-sourced level data is realized based on the model. Specifically, the following is described.
The embodiment provides a multi-contact advertisement attribution method. Fig. 1 is a flowchart of a multi-contact advertising attribution method according to an embodiment of the present application, as shown in fig. 1, the flowchart including the steps of:
a crowd-pack data set obtaining step S1, configured to obtain a data set including K crowd-packs, each crowd-pack corresponding to a conversion rate of a crowd-packEach crowd-sourced includes J ad spots, where K, J is a natural number, K e {1,2,..k }, J e {1,2,..j }. Specifically, each advertisement spot comprises a plurality of exposure records, and each exposure record comprises spot information and the number of exposure released on the spot.
Step S2 of data feature extraction, wherein feature extraction is carried out on each advertisement point in the data set to obtain feature vectors of each pointSpecifically, the point location information of each point location is subjected to feature coding based on the One-Hot coding method.
A model construction step S3 of constructing a semi-black box multi-contact attribution model based on crowd-sourced data to represent the relationship between each point location and crowd-sourced conversion rate, and based on feature vectorsDetermining model parameters;
specifically, the model construction step S3 further includes:
a relation model construction step S301, in which a semi-black box multi-contact attribution model for representing the relation between each point in the crowd-sourced and the crowd-sourced conversion rate is constructed, wherein the semi-black box multi-contact attribution model is as follows:
wherein,estimated conversion for the kth crowd-sourced,/->For the total exposure of the kth group of persons, n j Total number of exposure records for jth spot, +.>e (·) is by->Prediction b j W represents the network parameters of e (·), +.>The exposure quantity recorded for the ith exposure of the jth point of the k crowd pack;
step S302 of solving model parameters, constructing a loss function and calculating the loss function by using a gradient descent method to obtain network parameters w and parameters a j . Wherein the loss function is calculated from the following model:
wherein,estimated conversion for the kth crowd-sourced,/->Is the conversion rate of the kth crowd-sourced.
As above, the model construction step S3 provides a semi-black box multi-contact attribution model, the model realizes the prediction of the conversion rate of the advertisement point positions by collecting a plurality of groups of data training based on crowd package levels, effectively utilizes discrete and continuous characteristics, and jointly realizes the conversion contribution solving problem of the point positions by combining a neural network of a black box and a curve fitting method of a white box.
And S4, calculating the predicted contribution of each advertisement point in the data set based on the semi-black box multi-touch attribution model, and realizing advertisement attribution.
Based on the steps, the advertisement attribution method can be based on coarse-granularity level data, particularly crowd-sourced level data, advertisement attribution can be realized without specific point location data at a user level, and the business environment of the current higher data privacy protection requirement is met.
In some of these embodiments, the method further comprises:
and a model evaluation step S5, wherein the real contribution of each point location is acquired based on the data set, and the semi-black box multi-contact attribution model is evaluated according to the real contribution and the predicted contribution, so that the accuracy of the multi-contact advertisement attribution method is evaluated. Alternatively, R is used 2 The index evaluates a semi-black box multi-contact attribution model. Wherein the prediction contributionObtained from the following model calculations:
wherein,for the conversion exposure of the j-th spot, further: />
Based on the above, the present embodiment provides a method for predicting conversion contributions of a plurality of advertisement spots based on crowd-sourced data, which does not need to use user-level data and has high feasibility. In addition, the method fully utilizes the self characteristics of the advertisement points and the exposure characteristics of the advertisement points, thereby achieving a very good prediction effect.
The embodiments of the present application are described and illustrated below by means of preferred embodiments.
Fig. 4 is a flowchart of a multi-contact advertisement attribution method according to a preferred embodiment of the present application, as shown in fig. 4, which includes the steps of:
step S401: acquiring a data set containing K personal crowd-sourced data, wherein each crowd-sourced data corresponds to a crowd-sourced conversion rateK e {1, 2..k }. Each crowd-sourced data is composed of J advertisement points, each advertisement point comprises a plurality of exposure records, each record is composed of two parts, the first part is point information describing the point, and the second part is exposure quantity put on the corresponding point. By way of example and not limitation, the data format of a crowd-sourced group in the data set is shown in Table 1, the crowd-sourced group includes 6 fields, each row in Table 1 represents an exposure record, tpid is a point location identifier, media is a medium in which exposure is located, optional values of media include an archetype, tremble, etc., and action is exposureThe optional values of the device are the equipment where the exposure is located, and the optional values of the device are pc and moblie; the mark is the region where the exposure is located, the optional value of the mark is Beijing, shanghai, etc., and imp represents the exposure quantity under the constraint of media, action, device, market. The present embodiment defines a combination of media and action as a single point location. The data set provides a global true contribution y for each point j J e {1,2,., J }, can be used to evaluate algorithm accuracy.
Step S402: and extracting the characteristics of each point in the data set. Specifically, the feature extraction adopts a one-hot mode to encode media and actions, and the dimension imp is cascaded at the end after the encoding is completed. The format of each point position of the group packet data after the feature coding is finished is as follows:
wherein gp is k For the kth group of persons to be packaged,for the conversion of the kth crowd-sourced, < ->Is the eigenvector of point location j +.>For vectors encoded in one-hot for media, action +.>For imp value, J e {1,2,...
Step S403: and constructing an optimization model between the exposure of each point location and crowd-sourced conversion rate, namely a semi-black box multi-contact attribution model, and carrying out model solving. The construction process of the model is specifically as follows:
first, empirically we assume that the exposure and conversion exposure on a spot corresponds to the following functional relationship:
where x represents exposure, f (x) represents conversion exposure, a and b are parameters to be fitted, a is the upper bound of f (x), b controls the growth speed of f (x), and when a is the same, b is larger, and f (x) changes more slowly. This function is shown in fig. 5 by the dashed curve, from which it is seen that there will be an upper bound for f (x) with increasing x, i.e. for the converted exposure with increasing number of exposures. The straight line in the figure is to illustrate that at any one data point, the conversion exposure is less than the exposure. The above phenomena are consistent with the logic of exposure dose and conversion.
Then, considering that the properties of each point are different, for example, 100 exposures are put on points under the constraints of an Airy technology and a screen opening condition, and conversion exposure generated by 100 exposures are necessarily different when 100 exposures are put on points under the constraints of a tentative condition, we assume that each exposure point J epsilon {1, 2..the J } has an exposure-conversion exposure curve of itself, so as to construct a semi-black box multi-contact attribution model of the relationship between each point in a crowd pack and crowd pack conversion rate as follows:
wherein the method comprises the steps ofRepresenting the estimated conversion of the kth crowd-sourced,/->Indicating the total exposure of the kth group of persons,
the difference of the properties of each point in the model is mainly represented by conversion exposure brought by unit exposureThe larger the value, the better the point placement effect is proved. It is known that f (x) is controlled mainly in dependence of parameter b>The parameter b determines the nature of the point location, which is a variable closely related to the point location. Due to the feature vector +_in step S402>Embody the difference of the point positions, therefore +.>Prediction b j . At the same time, because the neural network has stronger prediction capability, we let +.>Where e (·) represents the neural network and w represents the network parameters.
FIG. 6 is a diagram of a semi-black box multi-contact attribution model organization structure according to a preferred embodiment of the present application, as can be seen with reference to FIG. 6, the model has two inputsAnd->Parameter a j Obtained by fitting f (x), parameter b j And predicting by a neural network.
Further, to solve for the unknown parameters w and a j The following loss function is constructedAnd calculating the loss function by using a gradient descent method to obtain a network parameter w and a parameter a j . Loss function->Defined as the following model:
s.t.a j ＞0
b j ＞0
wherein the method comprises the steps ofRepresenting the conversion rate of each crowd pack provided in the dataset,/->Indicating the total exposure on the j-th spot, < >>Loss function->The solved pseudocode is shown with reference to fig. 7.
Solving the steps to obtain parameters w and a j After that, we can calculate the conversion exposure at each spot, the conversion exposure at the j-th spot is:
the predicted contribution of each point location is:advertisement attribution is realized.
Step S404: to evaluate the accuracy of the model, the true contribution y on each point provided in the data set is determined j The above-mentioned predictive contributionBy R 2 The predicted contribution at each point of the index calculation is compared to the actual contribution provided in the dataset.
It should be noted that the steps illustrated in the above-described flow or flow diagrams of the figures may be performed in a computer system, such as a set of computer-executable instructions, and that although a logical order is illustrated in the flow diagrams, in some cases the steps illustrated or described may be performed in a different order than that illustrated herein.
The embodiment also provides a multi-contact advertisement attribution system, which is used for realizing the embodiment and the preferred implementation, and is not described in detail. As used below, the terms "module," "unit," "sub-unit," and the like may be a combination of software and/or hardware that implements a predetermined function. While the system described in the following embodiments is preferably implemented in software, implementation in hardware, or a combination of software and hardware, is also possible and contemplated.
FIG. 2 is a block diagram of a multi-contact advertising attribution system, as shown in FIG. 2, according to an embodiment of the present application, comprising:
a crowd-pack data set acquisition module 1 for acquiring a data set including K crowd-packs, each crowd-pack corresponding to a conversion rate of a crowd-packEach crowd-sourced includes J ad spots. Specifically, each advertisement spot comprises a plurality of exposure records, and each exposure record comprises a spotInformation and the number of exposures to be placed on the spot. Wherein K, J is a natural number, K is {1,2, K, J e {1, 2.
The data feature extraction module 2 performs feature extraction on each advertisement point in the data set to obtain feature vectors of each pointSpecifically, the point location information of each point location is subjected to feature coding based on the One-Hot coding method.
Model construction module 3 for constructing a semi-black box multi-contact attribution model based on crowd-sourced data to represent the relationship between each point and crowd-sourced conversion rate, and based on feature vectorsDetermining model parameters; wherein the model building module 3 further comprises: the relationship model construction module 301 constructs a semi-black box multi-contact attribution model for representing the relationship between each point in the crowd-sourced and the crowd-sourced conversion rate, wherein the semi-black box multi-contact attribution model is as follows:
wherein,estimated conversion for the kth crowd-sourced,/->For the total exposure of the kth group of persons, n j Total number of exposure records for jth spot, +.>e (·) is by->Prediction b j W represents the network parameters of e (·), +.>The exposure quantity recorded for the ith exposure of the jth point of the k crowd pack; model parameter solving module 302 constructs a loss function and calculates the loss function to obtain network parameters w and a parameters a by using gradient descent method j . Wherein the loss function is calculated from the following model:
wherein,estimated conversion for the kth crowd-sourced,/->Is the conversion rate of the kth crowd-sourced.
The semi-black box multi-contact attribution model obtained by the model construction module 3 realizes the prediction of the conversion rate of the advertisement point positions by collecting a plurality of groups of data training based on crowd pack level, effectively utilizes discrete and continuous characteristics, and jointly realizes the conversion contribution solving problem of the point positions by combining a neural network of a black box and a curve fitting method of a white box.
And the advertisement attribution module 4 calculates the prediction contribution of each advertisement point in the data set based on the semi-black box multi-contact attribution model to realize advertisement attribution.
Based on the module, the advertisement attribution method can be based on coarse-granularity level data, particularly crowd-sourced level data, advertisement attribution can be realized without specific point location data at a user level, and the business environment of the current higher data privacy protection requirement is met.
FIG. 3 is a block diagram of a preferred architecture of a multi-contact advertising attribution system, according to an embodiment of the present application, as shown in FIG. 3, the apparatus comprising all of the modules shown in FIG. 2, further comprising:
model evaluation module 5, baseThe true contribution of each spot bit is obtained from the dataset and a semi-black box multi-contact attribution model is evaluated based on the true contribution and the predicted contribution, thereby evaluating the accuracy of the multi-contact advertising attribution method. Alternatively, R is used 2 The index evaluates a semi-black box multi-contact attribution model. Wherein the predicted contribution is calculated from the following model:
wherein,conversion exposure for the j-th spot, specifically: />
The accuracy of the model can be effectively evaluated through the model evaluation module, so that the model can be conveniently optimized or adjusted.
The above-described respective modules may be functional modules or program modules, and may be implemented by software or hardware. For modules implemented in hardware, the various modules described above may be located in the same processor; or the above modules may be located in different processors in any combination.
In addition, the multi-contact advertising attribution method of the embodiments of the present application described in connection with fig. 1 may be implemented by a computer device. The computer device may include a processor and a memory storing computer program instructions.
In particular, the processor may include a Central Processing Unit (CPU), or an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), or may be configured to implement one or more integrated circuits of embodiments of the present application.
The memory may include, among other things, mass storage for data or instructions. By way of example, and not limitation, the memory may comprise a Hard Disk Drive (HDD), floppy Disk Drive, solid state Drive (Solid State Drive, SSD), flash memory, optical Disk, magneto-optical Disk, magnetic tape, or universal serial bus (Universal Serial Bus, USB) Drive, or a combination of two or more of these. The memory may include removable or non-removable (or fixed) media, where appropriate. The memory may be internal or external to the data processing apparatus, where appropriate. In a particular embodiment, the memory is a Non-Volatile (Non-Volatile) memory. In a particular embodiment, the Memory includes Read-Only Memory (ROM) and random access Memory (Random Access Memory, RAM). The ROM may be mask-programmed ROM, programmable ROM (Programmable Read-Only Memory, PROM), erasable PROM (Erasable Programmable Read-Only Memory, EPROM), electrically erasable PROM (Electrically Erasable Programmable Read-Only Memory, EEPROM), electrically rewritable ROM (Electrically Alterable Read-Only Memory, EAROM), or FLASH Memory (FLASH), or a combination of two or more of these, where appropriate. The RAM may be Static Random-Access Memory (SRAM) or dynamic Random-Access Memory (Dynamic RandomAccess Memory, DRAM) where the DRAM may be flash-mode dynamic Random-Access Memory (Fast Page Mode Dynamic Random Access Memory, FPMDRAM), extended data output dynamic Random-Access Memory (Extended Date Out Dynamic Random Access Memory, EDODRAM), synchronous dynamic Random-Access Memory (Synchronous Dynamic Random-Access Memory, SDRAM) or the like, as appropriate.
The memory may be used to store or cache various data files that need to be processed and/or communicated, as well as possible computer program instructions for execution by the processor.
The processor implements any of the multi-contact advertising attribution methods of the above embodiments by reading and executing computer program instructions stored in memory.
In addition, in combination with the multi-contact advertisement attribution method in the above embodiments, the embodiments of the present application may be implemented by providing a computer-readable storage medium. The computer readable storage medium has stored thereon computer program instructions; the computer program instructions, when executed by a processor, implement any of the multi-contact advertising attribution methods of the embodiments described above.
The technical features of the above-described embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above-described embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples merely represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.

Claims (8)

1. A multi-contact advertising attribution method, comprising:
a crowd pack data set acquisition step for acquiring a crowd pack data set includingKData sets of individual crowd-packs, each of the crowd-packs corresponding to a conversion rate of a crowd-packEach group package comprisesJAdvertisement points;
a data feature extraction step, wherein feature extraction is carried out on each advertisement point in the data set to obtain feature vectors of each point；
A model construction step of constructing a semi-black box multi-contact attribution model based on crowd-sourced data to represent the relationship between each point location and crowd-sourced conversion rate, and based on the feature vectorDetermining model parameters;
an advertisement attribution step, namely calculating the prediction contribution of each advertisement point in the data set based on the semi-black box multi-contact attribution model to realize advertisement attribution;
wherein,K、Jis a natural number of the Chinese characters,the model building step further includes:
a relation model construction step, namely constructing a half-black box multi-contact attribution model for representing the relation between each point in the crowd-sourced and the crowd-sourced conversion rate, wherein the half-black box multi-contact attribution model is as follows:
wherein,is->Estimated conversion rate of individual crowd-sourcing, +.>Is->Total exposure of individual crowd-sourcing,/->Is the firstjTotal number of exposure records of individual spots +.>，/>For passing->Prediction->Is->Representation ofNetwork parameters of->Is->Personal crowd pack->The first point of the dot positioniThe number of exposures recorded for each exposure;
model parameter solving step, constructing a loss function and calculating the loss function by using a gradient descent method to obtain the network parametersParameter->。
2. The multi-contact advertising attribution method as claimed in claim 1, further comprising:
and a model evaluation step of acquiring the real contribution of each point bit based on the data set and evaluating the semi-black box multi-contact attribution model according to the real contribution and the predicted contribution.
3. The multi-contact advertising attribution method according to claim 1, wherein the predicted contribution is calculated from the following model:
wherein,is the firstjConversion exposure of individual spots.
4. The multi-contact advertising attribution method according to claim 1, wherein the loss function is calculated from the following model:
wherein,is->Estimated conversion rate of individual crowd-sourcing, +.>Is->Conversion rate of individual crowd-sourcing.
5. A multi-contact advertising attribution system for performing the multi-contact advertising attribution method of any of claims 1-4, comprising:
crowd-sourced data set acquisition module for acquiring data includingKData sets of individual crowd-packs, each of the crowd-packs corresponding to a conversion rate of a crowd-packEach group package comprisesJAdvertisement points;
the data feature extraction module is used for carrying out feature extraction on each advertisement point in the data set to obtain feature vectors of each point；
The model construction module is used for constructing a semi-black box multi-contact attribution model based on crowd-sourced data so as to represent the relation between each point position and crowd-sourced conversion rate and based on the feature vectorDetermining model parameters;
the advertisement attribution module is used for calculating the prediction contribution of each advertisement point in the data set based on the semi-black box multi-contact attribution model to realize advertisement attribution;
wherein,K、Jis a natural number of the Chinese characters,the model building module further includes:
the relation model building module is used for building a half-black box multi-contact attribution model for representing the relation between each point in the crowd-sourced and the crowd-sourced conversion rate, and the half-black box multi-contact attribution model is as follows:
wherein,is->Estimated conversion rate of individual crowd-sourcing, +.>Is->Total exposure of individual crowd-sourcing,/->Is the firstjTotal number of exposure records of individual spots +.>，/>For passing->Prediction->Is->Representation ofNetwork parameters of->Is->Personal crowd pack->The first point of the dot positioniThe number of exposures recorded for each exposure;
model parameter solving module, constructing a loss function and calculating the loss function by using gradient descent method to obtain the network parameterParameter->。
6. The multi-contact advertising attribution system of claim 5, further comprising:
the model evaluation module acquires the real contribution of each point bit based on the data set and evaluates the semi-black box multi-contact attribution model according to the real contribution and the predicted contribution.
7. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the multi-contact advertising attribution method of any of claims 1-4 when the computer program is executed.
8. A computer-readable storage medium, on which a computer program is stored, characterized in that the program, when executed by a processor, implements a multi-contact advertising attribution method as claimed in any one of claims 1 to 4.
CN202110713526.4A 2021-06-25 2021-06-25 Multi-contact advertisement attribution method, system, computer device, and readable storage medium Active CN113450146B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110713526.4A CN113450146B (en) 2021-06-25 2021-06-25 Multi-contact advertisement attribution method, system, computer device, and readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110713526.4A CN113450146B (en) 2021-06-25 2021-06-25 Multi-contact advertisement attribution method, system, computer device, and readable storage medium

Publications (2)

Publication Number Publication Date
CN113450146A CN113450146A (en) 2021-09-28
CN113450146B true CN113450146B (en) 2024-02-02

Family

ID=77812940

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110713526.4A Active CN113450146B (en) 2021-06-25 2021-06-25 Multi-contact advertisement attribution method, system, computer device, and readable storage medium

Country Status (1)

Country Link
CN (1) CN113450146B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114693361A (en) * 2022-03-31 2022-07-01 北京明略软件系统有限公司 Method, device, equipment and storage medium for determining contribution degree of channel

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110147882A (en) * 2018-09-03 2019-08-20 腾讯科技（深圳）有限公司 Training method, crowd's method of diffusion, device and the equipment of neural network model
CN110796499A (en) * 2019-11-06 2020-02-14 中山大学 Advertisement conversion rate estimation model and training method thereof
CN111311314A (en) * 2020-01-21 2020-06-19 华为技术有限公司 Advertisement attribution method and equipment
CN112529634A (en) * 2020-12-18 2021-03-19 恩亿科(北京)数据科技有限公司 Transformation link analysis method and system based on big data and computer equipment

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11287894B2 (en) * 2018-03-09 2022-03-29 Adobe Inc. Utilizing a touchpoint attribution attention neural network to identify significant touchpoints and measure touchpoint contribution in multichannel, multi-touch digital content campaigns

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110147882A (en) * 2018-09-03 2019-08-20 腾讯科技（深圳）有限公司 Training method, crowd's method of diffusion, device and the equipment of neural network model
CN110796499A (en) * 2019-11-06 2020-02-14 中山大学 Advertisement conversion rate estimation model and training method thereof
CN111311314A (en) * 2020-01-21 2020-06-19 华为技术有限公司 Advertisement attribution method and equipment
CN112529634A (en) * 2020-12-18 2021-03-19 恩亿科(北京)数据科技有限公司 Transformation link analysis method and system based on big data and computer equipment

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于改进Wide＆Deep 交互特征提取的移动APP 转化率预估;孙晓燕;《郑州大学学报( 工学版)》;第26-32页 *

Also Published As

Publication number Publication date
CN113450146A (en) 2021-09-28

Similar Documents

Publication Publication Date Title
CN108875916B (en) Advertisement click rate prediction method based on GRU neural network
Smith et al. Estimation of copula models with discrete margins via Bayesian data augmentation
US10565525B2 (en) Collaborative filtering method, apparatus, server and storage medium in combination with time factor
CN109919316A (en) The method, apparatus and equipment and storage medium of acquisition network representation study vector
CN110210233A (en) Joint mapping method, apparatus, storage medium and the computer equipment of prediction model
CN106909560B (en) Interest point ordering method
CN110689110B (en) Method and device for processing interaction event
WO2015175835A1 (en) Click through ratio estimation model
CN113450146B (en) Multi-contact advertisement attribution method, system, computer device, and readable storage medium
CN106919564A (en) A kind of influence power measure based on mobile subscriber&#39;s behavior
Xu et al. Predicting advertiser bidding behaviors in sponsored search by rationality modeling
CN111861178B (en) Training method of service matching model, service matching method, equipment and medium
Inoue et al. Estimating customer impatience in a service system with unobserved balking
CN112561569B (en) Dual-model-based store arrival prediction method, system, electronic equipment and storage medium
CN116361643A (en) Model training method for realizing object recommendation, object recommendation method and related device
Lin et al. Construction of disease risk scoring systems using logistic group lasso: application to porcine reproductive and respiratory syndrome survey data
Zhao et al. Extreme tail risk estimation with the generalized Pareto distribution under the peaks-over-threshold framework
CN114463590A (en) Information processing method, apparatus, device, storage medium, and program product
CN115345635A (en) Processing method and device for recommended content, computer equipment and storage medium
CN113781110A (en) User behavior prediction method and system based on multi-factor weighted BI-LSTM learning
WO2016041376A1 (en) Method and device for predicting information propagation in social network
US20210334311A1 (en) Method, apparatus, device and storage medium for determining a central vertex in a social network
Faridani et al. Combining trust in collaborative filtering to mitigate data sparsity and cold-start problems
CN113676505B (en) Information pushing method, device, computer equipment and storage medium
Alshahrani et al. Influence maximization based global structural properties: A multi-armed bandit approach

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
GR01 Patent grant
GR01 Patent grant