CN-115471271-B - Advertisement attribution method, advertisement attribution device, computer equipment and readable storage medium

CN115471271BCN 115471271 BCN115471271 BCN 115471271BCN-115471271-B

Abstract

The invention discloses an advertisement attribution method, a device, a computer device and a readable storage medium, relating to the attribution field, wherein the method comprises the steps of obtaining user data and advertisement path data; the method comprises the steps of encoding identification data of each contact according to an encoding module of an attribution model to obtain identification data representation of each contact, generating weight of each contact by using an attention module of the attribution model, obtaining path representation of advertisement path data based on the identification data representation and the weight of each contact, inputting user data and channel data of a plurality of contacts to a second acquisition module of the attribution model to obtain data characteristic representation, and inputting the path representation and the data characteristic representation to a prediction module of the attribution model to obtain conversion probability. Therefore, the invention ensures that the prediction of the conversion probability considers the influence of different user trigger time on the contribution degree of the contact, so that the contribution degree of the contact can be related to the time factor of the user accessing the advertisement, and the conversion probability is ensured to be effective.

Inventors

SUN YI
WANG SHUYUE
LIN ZHAOWEN
XU TONG

Assignees

北京邮电大学

Dates

Publication Date: 20260508
Application Date: 20220929

Claims (10)

1. A method for attributing advertisements, comprising: Acquiring user data and advertisement path data, wherein the advertisement path data comprises contact point data of a plurality of contact points, the contact point data comprises identification data and channel data, and the contact point data of the plurality of contact points are ordered according to user triggering time; coding the identification data of each contact based on a coding module of a pre-trained attribution model to obtain the identification data representation of each contact; generating a weight of each contact based on the attention module of the attribution model, wherein the greater the difference between the user trigger time of the contact and the user trigger time of the last contact, the lower the weight of the contact; obtaining a path representation of the advertisement path data based on the identification data representation and the weight of each contact; Inputting the user data and channel data of the contacts to a second acquisition module of the attribution model to obtain data characteristic representation, wherein the second acquisition module is used for respectively embedding the user data and the channel data of the contacts to obtain corresponding embedded vectors, and outputting the data characteristic representation after all the embedded vectors are spliced; And inputting the path representation and the data characteristic representation to a prediction module of the attribution model to obtain conversion probability.
2. The attribution method of an advertisement according to claim 1, wherein the attribution model further comprises a splice module, a flat module, and a full connection module; the obtaining the path representation of the advertisement path data based on the identification data representation and the weight of each contact comprises the following steps: calculating the product of the identification data representation and the weight of the contact to obtain the weighted data representation of the contact; Inputting the weighted data representation of each contact to the splicing module to obtain a first splicing result; Inputting the first splicing result to the flat module to obtain a flattening result; And inputting the flattening result to the full-connection module to obtain the path representation of the advertisement path data.
3. The attribution method of an advertisement according to claim 1, wherein the second obtaining module comprises an embedded layer, a first splicing layer, and a first full connection layer; the inputting the user data and channel data of the contacts to a second acquisition module of the attribution model to obtain a data characteristic representation, including: Respectively inputting the user data and channel data of the contacts to the embedding layer to obtain an embedding vector corresponding to the user data and an embedding vector corresponding to the channel data of each contact; Inputting all the embedded vectors into the first splicing layer to obtain a second splicing result; and inputting all second splicing results into the first full-connection layer to obtain data characteristic representation.
4. The attribution method of an advertisement according to claim 1, wherein the prediction module comprises a second splicing layer, a second full connection layer, and a probability calculation layer; The inputting the path representation and the data feature representation into a prediction module of the attribution model, obtaining conversion probability, comprising: inputting the path representation and the data characteristic representation to the second splicing layer to obtain a third splicing result; Inputting the third splicing result to the second full-connection layer to obtain corresponding output data; and inputting the output data to the probability calculation layer to obtain conversion probability.
5. The method of attribution of advertisements of claim 1, wherein the obtaining user data and advertisement path data comprises: acquiring original user data and original advertisement path data; and carrying out data cleaning on the original user data and the original advertisement path data to obtain user data and advertisement path data.
6. The method of attribution of advertisements of claim 1, wherein the encoding module is implemented based on a two-way long and short term memory network.
7. An advertising attribution device, comprising: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for acquiring user data and advertisement path data, the advertisement path data comprises contact point data of a plurality of contact points, the contact point data comprises identification data and channel data, and the contact point data of the plurality of contact points are ordered according to user triggering time; The coding module is used for coding the identification data of each contact based on the pre-trained coding module of the attribution model to obtain the identification data representation of each contact; The generation module is used for generating the weight of each contact based on the attention module of the attribution model, wherein the larger the difference between the user trigger time of the contact and the user trigger time of the last contact is, the lower the weight of the contact is; a path representation acquisition module, configured to obtain a path representation of the advertisement path data based on the identification data representation and the weight of each contact; The feature representation acquisition module is used for inputting the user data and the channel data of the contacts to the second acquisition module of the attribution model to obtain a data feature representation, wherein the second acquisition module is used for respectively embedding the user data and the channel data of the contacts to obtain corresponding embedded vectors, and outputting the data feature representation after all the embedded vectors are spliced; and the prediction module is used for inputting the path representation and the data characteristic representation into the prediction module of the attribution model to obtain conversion probability.
8. The advertising attribution device of claim 7, wherein the attribution model further comprises a stitching module, a flat module, and a fully connected module; The path representation acquisition module includes: The product calculation sub-module is used for calculating the product of the identification data representation and the weight of the contact to obtain the weighted data representation of the contact; the first splicing sub-module is used for inputting the weighted data representation of each contact to the splicing module to obtain a first splicing result; The first flattening processing submodule is used for inputting the first splicing result to the flattening module to obtain a flattening result; And the full-connection processing sub-module is used for inputting the flattening result to the full-connection module to obtain the path representation of the advertisement path data.
9. A computer device comprising a memory and a processor, the memory storing a computer program that, when run on the processor, performs the attribution method of an advertisement of any of claims 1-6.
10. A computer readable storage medium, having stored thereon a computer program, which when run on a processor performs the attribution method of an advertisement as claimed in any of claims 1-6.

Description

Advertisement attribution method, advertisement attribution device, computer equipment and readable storage medium Technical Field The present invention relates to attribution, and more particularly, to an advertisement attribution method, apparatus, computer device, and readable storage medium. Background Common attribution methods are primary contact attribution, secondary interaction attribution and linear attribution. Wherein the initial de-attribution and the last interaction attribution attribute all contributions of the conversion to a first contact or a last contact in the user's advertising campaign path. Linear attribution is then the uniform distribution of the translated contribution to each contact in the ad campaign path. It is clear that the first contact neglects the presentation/exposure effect of each contact before conversion, the last interaction neglects the difference in contribution of the different contacts due to the linear attribution due to overestimated contribution of the last contact. In other words, it is difficult for the existing advertisement attribution method to reasonably/correctly embody the connection between the conversion and the contact, so that the advertisement attribution effect/precision is poor. Disclosure of Invention In view of the above, the present invention provides an advertisement attribution method, apparatus, computer device and readable storage medium, which are used for improving the current situation that the existing advertisement attribution method is difficult to reasonably/correctly embody the connection between conversion and contact, so that the attribution effect/precision of advertisement is poor. In a first aspect, an embodiment of the present invention provides an attribution method for an advertisement, including: Acquiring user data and advertisement path data, wherein the advertisement path data comprises contact point data of a plurality of contact points, the contact point data comprises identification data and channel data, and the contact point data of the plurality of contact points are ordered according to user triggering time; coding the identification data of each contact based on a coding module of a pre-trained attribution model to obtain the identification data representation of each contact; generating a weight of each contact based on the attention module of the attribution model, wherein the greater the difference between the user trigger time of the contact and the user trigger time of the last contact, the lower the weight of the contact; obtaining a path representation of the advertisement path data based on the identification data representation and the weight of each contact; Inputting the user data and channel data of the contacts to a second acquisition module of the attribution model to obtain data characteristic representation, wherein the second acquisition module is used for respectively embedding the user data and the channel data of the contacts to obtain corresponding embedded vectors, and outputting the data characteristic representation after all the embedded vectors are spliced; And inputting the path representation and the data characteristic representation to a prediction module of the attribution model to obtain conversion probability. Optionally, in a feasible manner provided by the embodiment of the present invention, the attribution model further includes a splicing module, a flat module and a full connection module; the obtaining the path representation of the advertisement path data based on the identification data representation and the weight of each contact comprises the following steps: calculating the product of the identification data representation and the weight of the contact to obtain the weighted data representation of the contact; Inputting the weighted data representation of each contact to the splicing module to obtain a first splicing result; Inputting the first splicing result to the flat module to obtain a flattening result; And inputting the flattening result to the full-connection module to obtain the path representation of the advertisement path data. Optionally, in a feasible manner provided by the embodiment of the present invention, the second obtaining module includes an embedded layer, a first splicing layer and a first full connection layer; the inputting the user data and channel data of the contacts to a second acquisition module of the attribution model to obtain a data characteristic representation, including: Respectively inputting the user data and channel data of the contacts to the embedding layer to obtain an embedding vector corresponding to the user data and an embedding vector corresponding to the channel data of each contact; Inputting all the embedded vectors into the first splicing layer to obtain a second splicing result; and inputting all second splicing results into the first full-connection layer to obtain data characteristic representation. Optionally,