CN-122022921-A - Advertisement plan delivery method and related device

Abstract

The application provides a method and a related device for delivering an advertisement plan. The method comprises the steps of generating model prediction conversion rate of an advertisement display offer sent by an advertisement platform and related to a target advertisement plan by using a specified prediction model, performing first-stage calibration on the model prediction conversion rate according to a pre-constructed cross-domain mapping relation to obtain corrected prediction conversion rate, wherein the cross-domain mapping relation is used for representing a mapping relation between an output result space of the specified prediction model and a historical conversion rate distribution space of the advertisement platform, and performing second-stage calibration on the corrected prediction conversion rate according to natural in-store conversion rate of a store corresponding to the target advertisement plan to obtain the target prediction conversion rate. The application can realize that the artificial intelligent model based on the training of the small sample is applied to the advertising platform.

Inventors

• LIU YAQI
• GAO LONGYU
• HUANG PINGCHUN

Assignees

• 阿里健康科技(中国)有限公司

Dates

Publication Date

20260512

Application Date

20251230

Claims (10)

1. 1. A method for delivering an advertising program, comprising: Generating model predictive conversion rate of the advertisement display offers by using a specified predictive model aiming at the advertisement display offers which are sent by an advertisement platform and are associated with a target advertisement plan, wherein the model predictive conversion rate is used for representing the probability of facilitating commodity orders after commodity advertisement information corresponding to the target advertisement plan is displayed; Performing first-stage calibration on the model predictive conversion rate according to a pre-constructed cross-domain mapping relation to obtain a corrected predictive conversion rate, wherein the cross-domain mapping relation is used for representing a mapping relation between an output result space of the appointed predictive model and a historical conversion rate distribution space of the advertisement platform; and carrying out second-stage calibration on the corrected predicted conversion rate according to the natural business-in conversion rate of the store corresponding to the target advertisement plan to obtain the target predicted conversion rate, wherein the natural business-in conversion rate is used for representing the historical conversion rate distribution of the store issuing the target advertisement plan in a natural business-in scene.
2. 2. The method according to claim 1, wherein the method further comprises: the method comprises the steps of obtaining historical advertisement display data in an advertisement platform, wherein the historical advertisement display data comprises a plurality of advertisement samples, and the advertisement samples comprise displayed commodity advertisement information and corresponding display results; Performing equal-frequency barrel dividing processing on the advertisement samples to obtain a plurality of barrels; calculating the actual conversion rate of each sub-bucket according to the display result of the advertisement sample in each sub-bucket; and constructing the cross-domain mapping relation according to the model prediction conversion rate and the actual conversion rate of the specified prediction model corresponding to the advertisement sample.
3. 3. The method of claim 2, wherein constructing the cross-domain mapping relationship from the model predicted conversion rate and the actual conversion rate of the specified prediction model corresponding to the advertisement samples comprises: And determining a model prediction conversion rate interval corresponding to the advertisement sample in each sub-bucket, wherein the cross-domain mapping relation comprises a corresponding model medical conversion rate interval and an actual conversion rate.
4. 4. The method of claim 2, wherein the historical advertisement presentation data comprises a sample set and a verification set, wherein the sample set comprises the plurality of advertisement samples; According to the model prediction conversion rate and the actual conversion rate of the specified prediction model corresponding to the advertisement sample, constructing the cross-domain mapping relation comprises the following steps: Determining a sub-bucket corresponding to each verification sample; generating an actual conversion rate of the verification sample in each sub-bucket; and determining the cross-domain mapping relation according to the actual conversion rate of the advertisement samples in the multiple sub-buckets and the actual conversion rate of the verification samples.
5. 5. The method of claim 1, further comprising recalling a health requirement tag of the advertisement presentation offer corresponding to the user, wherein the health requirement tag of the user is set based on behavioral data of the user; and determining an associated advertisement plan as the target advertisement plan according to the health requirement label, wherein the commodity corresponding to the target advertisement plan has the same health requirement label as the user.
6. 6. The method of claim 5, wherein the behavioral data comprises merchandise behavioral data generated by a user's operational behaviors with respect to a plurality of merchandise; the method further comprises the steps of: Generating a potential user interest code according to commodity semantic codes of commodities, wherein the commodity semantic codes are constructed based on commodity text information of the corresponding commodities; And under the condition that the user interest code of the user aimed at by the advertisement showing offer is matched with the user interest code, using the advertisement showing offer as the advertisement showing offer associated with the target advertisement plan of the commodity.
7. 7. The method of claim 6, wherein the method further comprises: vectorizing commodity behavior data generated in the commodity browsing process of a user to obtain commodity behavior vectorization representation; Carrying out pooling treatment on commodity behavior vectorization representation corresponding to a user to obtain user vector representation of the user; And calling a residual quantization variation self-encoder according to the user vector representation to generate the user interest code of the user.
8. 8. A computer readable storage medium, having stored thereon a computer program which, when executed by a processor, causes the processor to implement the method of any of claims 1 to 7.
9. 9. A computer device comprising a memory and a processor, wherein the memory stores at least one computer program, the at least one computer program being loaded and executed by the processor to implement the method of any one of claims 1 to 7.
10. 10. A computer program product comprising computer instructions which, when executed by a processor, implement the method of any of claims 1 to 7.

Description

Advertisement plan delivery method and related device Technical Field One or more embodiments of the present application relate to the field of artificial intelligence technologies, and in particular, to a method and an apparatus for delivering an advertisement plan. Background In advertising for internet medical e-commerce, a conversion rate prediction model is typically trained based on historical exposure, click-through and conversion data, and bidding control is performed by a computer automatically employing the predicted conversion rate output by the model upon receipt of an advertisement presentation offer. However, for cold start advertising programs or less sample size advertising programs, the conversion predictive model is difficult to train effectively due to the lack of sufficient training samples. When the conversion rate prediction model is applied to the advertisement platform, abnormal delivery of the advertisement plan can be caused, and normal use of the advertisement platform can not be met. Disclosure of Invention In view of this, one or more embodiments of the present application provide a method and related apparatus for delivering an advertisement plan, which can implement an artificial intelligence model based on small sample training to be applied to an advertisement platform. According to one or more embodiments of the application, a method for delivering an advertisement plan is provided, wherein the method comprises the steps of generating model predictive conversion rate of an advertisement presentation offer sent by an advertisement platform and associated with a target advertisement plan by using a specified predictive model, wherein the model predictive conversion rate is used for representing probability of promoting commodity orders after commodity advertisement information corresponding to the target advertisement plan is presented, performing first-stage calibration on the model predictive conversion rate according to a pre-constructed cross-domain mapping relation to obtain corrected predictive conversion rate, wherein the cross-domain mapping relation is used for representing a mapping relation between an output result space of the specified predictive model and a historical conversion rate distribution space of the advertisement platform, and performing second-stage calibration on the corrected predictive conversion rate according to a natural business-entering conversion rate of a store corresponding to the target advertisement plan to obtain target predictive conversion rate, wherein the natural business-entering conversion rate is used for representing a historical conversion rate distribution of the store issuing the target advertisement plan under a natural business-entering scene. According to the second aspect, one or more embodiments of the application provide a device for delivering an advertisement plan, which comprises a generation module, a first calibration module and a second calibration module, wherein the generation module is used for generating a model prediction conversion rate of an advertisement display offer sent by an advertisement platform and related to a target advertisement plan by using a specified prediction model, the model prediction conversion rate is used for representing the probability of promoting a commodity order after commodity advertisement information corresponding to the target advertisement plan is presented, the first calibration module is used for carrying out first-stage calibration on the model prediction conversion rate according to a pre-constructed cross-domain mapping relation to obtain a corrected prediction conversion rate, the cross-domain mapping relation is used for representing a mapping relation between an output result space of the specified prediction model and a historical conversion rate distribution space of the advertisement platform, and the second calibration module is used for carrying out second-stage calibration on the corrected prediction conversion rate according to a natural business-in conversion rate corresponding to a store, and the natural business-in conversion rate is used for representing the historical conversion rate of the store of the target advertisement plan under a natural business-in scene. In a third aspect, one or more embodiments of the present application provide a computer device, the computer device including a memory and a processor, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to implement a method as described above. In a fourth aspect, one or more embodiments of the application provide a computer program product comprising computer instructions which, when executed by a processor, implement a method as described above. In a fifth aspect, one or more embodiments of the present application provide a computer-readable storage medium having stored thereon a computer program which, when executed by