CN117132326A - Advertisement pushing method and device, electronic equipment and storage medium - Google Patents

Advertisement pushing method and device, electronic equipment and storage medium Download PDF

Info

Publication number
CN117132326A
CN117132326A CN202311085864.3A CN202311085864A CN117132326A CN 117132326 A CN117132326 A CN 117132326A CN 202311085864 A CN202311085864 A CN 202311085864A CN 117132326 A CN117132326 A CN 117132326A
Authority
CN
China
Prior art keywords
pushing
advertisement
training sample
model
push
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
CN202311085864.3A
Other languages
Chinese (zh)
Inventor
成楚璇
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.)
China Merchants Bank Co Ltd
Original Assignee
China Merchants Bank 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 China Merchants Bank Co Ltd filed Critical China Merchants Bank Co Ltd
Priority to CN202311085864.3A priority Critical patent/CN117132326A/en
Publication of CN117132326A publication Critical patent/CN117132326A/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
    • G06Q30/0241—Advertisements
    • G06Q30/0251—Targeted advertisements
    • G06Q30/0269—Targeted advertisements based on user profile or attribute
    • G06Q30/0271—Personalized advertisement
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • 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
    • G06N3/0475—Generative 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
    • G06Q30/0241—Advertisements
    • G06Q30/0242—Determining effectiveness of advertisements
    • 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
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

The application discloses an advertisement pushing method, an advertisement pushing device, electronic equipment and a storage medium, and relates to the field of advertisement pushing. Compared with the pushing method which only takes the click rate as a target and forcibly increases the exposure proportion of the related materials of the financial products in the traditional scheme, the embodiment pushes the advertisement to the user by combining the click probability and the purchase probability, thereby synchronously improving the click rate of the advertisement and the purchase rate of the products on the advertisement, namely improving the pushing return rate of the product advertisements.

Description

Advertisement pushing method and device, electronic equipment and storage medium
Technical Field
The present application relates to the field of advertisement pushing technologies, and in particular, to an advertisement pushing method, an advertisement pushing device, an electronic device, and a storage medium.
Background
In the banking and finance industry, the traditional pushing scheme can effectively and accurately conduct financial advertisement recommendation at present, and the viscosity of a user is improved, but the recommendation method is based on the aim of clicking rate, wherein the clicking rate is the number of clicking points/the number of exposure points, under the method, rights and benefits materials have exposure advantages, financial product related material exposure can be suppressed, in order to ensure the exposure proportion of financial product related materials, operators can forcedly limit the exposure duty ratio of other types of materials, and though the exposure rate of the financial product materials is improved, the overall clicking rate is reduced, so that the pushing return rate of the financial product advertisements is still lower.
Disclosure of Invention
The application mainly aims to provide an advertisement pushing method, an advertisement pushing device, electronic equipment and a storage medium, and aims to solve the technical problem that the current financial product advertisement pushing return rate is low.
In order to achieve the above object, the present application provides an advertisement pushing method, which includes:
extracting pushing characteristics from a pushing scene;
based on the pushing characteristics, predicting the purchase probability of the pushed product in the pushing scene and the click probability of the recommended advertisement of the pushed product;
based on the purchase probability and the click probability, predicting a push value for pushing the recommended advertisement to a push object in the push scene;
and pushing recommended advertisements to the pushing objects based on the pushing values.
Optionally, the pushing features include a user feature, an advertisement feature, an interaction feature, and a financial feature, and the step of predicting, based on the pushing features, a purchase probability of a pushed product in the pushing scene and a click probability of a recommended advertisement of the pushed product includes:
inputting the user characteristics, the advertisement characteristics and the interaction characteristics into a first prediction model for prediction to obtain click probability of the recommended advertisement;
And inputting the user characteristics, the advertisement characteristics, the interaction characteristics and the financial characteristics into a second prediction model to obtain the purchase probability of the pushed product.
Optionally, the step of predicting a push value of pushing the recommended advertisement to a push object in the push scene based on the purchase probability and the click probability includes:
and inputting the purchase probability and the click probability into a preset fusion task model to obtain a pushing value for pushing the recommended advertisement to the pushing object, wherein a first weight of the purchase probability and a second weight of the click probability in the preset fusion task model are generated through a self-adaptive weight generation network.
Optionally, before the step of predicting the purchase probability of the pushed product and the click probability of the recommended advertisement of the pushed product in the push scene based on the push feature, the method comprises:
the method comprises the steps of carrying out iterative training on an initialization model based on a preset training sample set to obtain a first prediction model and a second prediction model, wherein the types of training samples in the preset training sample set comprise a first training sample and a second training sample, the labels of the first training sample are clicked or not clicked, the labels of the second training sample are purchased or not purchased, the first prediction model is obtained through training of the first training sample, and the second prediction model is obtained through training of the second training sample.
Optionally, the initialization model includes a first initialization model and a second initialization model, and the step of iteratively training the initialization model based on a preset training sample set to obtain the first prediction model and the second prediction model includes:
for any first training sample, inputting the first training sample into the first initialization model to obtain a first prediction result of the first initialization model;
updating parameters in the first initialization model based on the difference between the first prediction result and the label of the first training sample, returning to execute the step of inputting the first training sample into the first initialization model based on the new first training sample to obtain a first prediction result of the first initialization model until a preset training completion condition is reached, and taking the first initialization model as the first prediction model;
for any second training sample, inputting the second training sample into the second initialization model to obtain a second prediction result of the second initialization model;
updating parameters in the second initialization model based on the difference between the second prediction result and the label of the second training sample, and returning to execute the step of inputting the second training sample into the second initialization model based on the new second training sample to obtain a second prediction result of the second initialization model until a preset training completion condition is reached, wherein the second initialization model is used as the second prediction model.
Optionally, before the step of inputting the first training sample into the first initialization model and/or before the step of inputting the second training sample into the second initialization model, the method comprises:
the first training sample is preprocessed, and/or the second training sample is preprocessed, wherein the preprocessing comprises null filling, and/or value mapping.
Optionally, the financial features in the push feature include a financial product feature and a financial product cross feature, wherein the financial product cross feature is determined based on data generated by a financial product cross with a user and/or advertising scene.
In addition, in order to achieve the above object, the present application also provides an advertisement pushing device, including:
the extraction module is used for extracting pushing characteristics from the pushing scene;
the first prediction module is used for predicting the purchase probability of the pushed product in the pushed scene and the click probability of the recommended advertisement of the pushed product based on the push characteristics;
the second prediction module is used for predicting a push value of pushing the recommended advertisement to a push object in the push scene based on the purchase probability and the click probability;
And the pushing module is used for pushing the recommended advertisement to the pushing object based on the pushing value.
In addition, to achieve the above object, the present application also provides an electronic device including: the advertisement pushing device comprises a memory, a processor and an advertisement pushing program which is stored in the memory and can run on the processor, wherein the advertisement pushing program realizes the steps of the advertisement pushing method when being executed by the processor.
In addition, in order to achieve the above object, the present application also provides a storage medium having stored thereon an advertisement pushing program which, when executed by a processor, implements the steps of the advertisement pushing method as described above.
The embodiment of the application provides an advertisement pushing method, an advertisement pushing device, electronic equipment and a storage medium. In the embodiment of the application, the pushing feature is extracted from the pushing scene; based on the pushing characteristics, predicting the purchase probability of the pushed product in the pushing scene and the click probability of the recommended advertisement of the pushed product; based on the purchase probability and the click probability, predicting a push value for pushing the recommended advertisement to a push object in the push scene; and pushing recommended advertisements to the pushing objects based on the pushing values. In other words, when pushing advertisements, the click probability of recommended advertisements and the purchase probability of the pushed products are respectively obtained through the pushing features extracted from the pushing scene, the click probability and the purchase probability are integrated, the pushing value of pushing the recommended advertisements to the users is obtained through comprehensive consideration, and the recommended advertisements are pushed to the users based on the pushing value. Compared with the pushing method which only takes the click rate as a target and forcibly increases the exposure proportion of the related materials of the financial products in the traditional scheme, the embodiment pushes the advertisement to the user by combining the click probability and the purchase probability, thereby synchronously improving the click rate of the advertisement and the purchase rate of the products on the advertisement, namely improving the pushing return rate of the product advertisements.
Drawings
FIG. 1 is a schematic diagram of an electronic device in a hardware operating environment according to an embodiment of the present application;
FIG. 2 is a flowchart illustrating a first embodiment of an advertisement pushing method according to the present application;
FIG. 3 is a schematic diagram of an overall model framework in the advertisement pushing method of the present application;
FIG. 4 is a flowchart illustrating a second embodiment of an advertisement pushing method according to the present application;
fig. 5 is a schematic diagram of an advertisement pushing device in the advertisement pushing method of the present application.
The achievement of the objects, functional features and advantages of the present application will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
Referring to fig. 1, fig. 1 is a schematic diagram of an electronic device structure of a hardware running environment according to an embodiment of the present application.
The electronic equipment of the embodiment of the application can be a server, and also can be electronic terminal equipment such as a smart phone, a PC, a tablet personal computer, a portable computer and the like.
As shown in fig. 1, the electronic device may include: a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, a communication bus 1002. Wherein the communication bus 1002 is used to enable connected communication between these components. The user interface 1003 may include a Display, an input unit such as a Keyboard (Keyboard), and the optional user interface 1003 may further include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface). The memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may also optionally be a storage device separate from the processor 1001 described above.
Optionally, the electronic device may further include a camera, an RF (Radio Frequency) circuit, a sensor, an audio circuit, a WiFi module, and the like. The terminal may also be configured with other sensors such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc., which are not described in detail herein. Those skilled in the art will appreciate that the electronic device structure shown in fig. 1 is not limiting of the electronic device and may include more or fewer components than shown, or may combine certain components, or may be arranged in different components.
Those skilled in the art will appreciate that the electronic device structure shown in fig. 1 is not limiting of the electronic device and may include more or fewer components than shown, or may combine certain components, or may be arranged in different components.
Further, as shown in fig. 1, an operating system, a network communication module, a user interface module, and an advertisement push program may be included in the memory 1005, which is a kind of computer storage medium.
In the electronic device shown in fig. 1, the network interface 1004 is mainly used for connecting to a background server and performing data communication with the background server; the user interface 1003 is mainly used for connecting a client (user side) and performing data communication with the client; and the processor 1001 may be configured to call an advertisement push program stored in the memory 1005 and perform the following operations:
Extracting pushing characteristics from a pushing scene;
based on the pushing characteristics, predicting the purchase probability of the pushed product in the pushing scene and the click probability of the recommended advertisement of the pushed product;
based on the purchase probability and the click probability, predicting a push value for pushing the recommended advertisement to a push object in the push scene;
and pushing recommended advertisements to the pushing objects based on the pushing values.
In a possible implementation, the processor 1001 may call an advertisement push program stored in the memory 1005, and further perform the following operations:
the pushing features comprise user features, advertisement features, interaction features and financial features, and the step of predicting the purchase probability of the pushed product and the click probability of the recommended advertisement of the pushed product in the pushing scene based on the pushing features comprises the following steps:
inputting the user characteristics, the advertisement characteristics and the interaction characteristics into a first prediction model for prediction to obtain click probability of the recommended advertisement;
and inputting the user characteristics, the advertisement characteristics, the interaction characteristics and the financial characteristics into a second prediction model to obtain the purchase probability of the pushed product.
In a possible implementation, the processor 1001 may call an advertisement push program stored in the memory 1005, and further perform the following operations:
the step of predicting a push value for pushing the recommended advertisement to a push object in the push scene based on the purchase probability and the click probability includes:
and inputting the purchase probability and the click probability into a preset fusion task model to obtain a pushing value for pushing the recommended advertisement to the pushing object, wherein a first weight of the purchase probability and a second weight of the click probability in the preset fusion task model are generated through a self-adaptive weight generation network.
In a possible implementation, the processor 1001 may call an advertisement push program stored in the memory 1005, and further perform the following operations:
before the step of predicting a purchase probability of the pushed product and a click probability of the recommended advertisement of the pushed product in the push scenario based on the push feature, the method comprises:
the method comprises the steps of carrying out iterative training on an initialization model based on a preset training sample set to obtain a first prediction model and a second prediction model, wherein the types of training samples in the preset training sample set comprise a first training sample and a second training sample, the labels of the first training sample are clicked or not clicked, the labels of the second training sample are purchased or not purchased, the first prediction model is obtained through training of the first training sample, and the second prediction model is obtained through training of the second training sample.
In a possible implementation, the processor 1001 may call an advertisement push program stored in the memory 1005, and further perform the following operations:
the initialization model comprises a first initialization model and a second initialization model, and the step of performing iterative training on the initialization model based on a preset training sample set to obtain the first prediction model and the second prediction model comprises the following steps:
for any first training sample, inputting the first training sample into the first initialization model to obtain a first prediction result of the first initialization model;
updating parameters in the first initialization model based on the difference between the first prediction result and the label of the first training sample, returning to execute the step of inputting the first training sample into the first initialization model based on the new first training sample to obtain a first prediction result of the first initialization model until a preset training completion condition is reached, and taking the first initialization model as the first prediction model;
for any second training sample, inputting the second training sample into the second initialization model to obtain a second prediction result of the second initialization model;
Updating parameters in the second initialization model based on the difference between the second prediction result and the label of the second training sample, and returning to execute the step of inputting the second training sample into the second initialization model based on the new second training sample to obtain a second prediction result of the second initialization model until a preset training completion condition is reached, wherein the second initialization model is used as the second prediction model.
In a possible implementation, the processor 1001 may call an advertisement push program stored in the memory 1005, and further perform the following operations:
before the step of inputting the first training sample into the first initialization model and/or before the step of inputting the second training sample into the second initialization model, the method comprises:
the first training sample is preprocessed, and/or the second training sample is preprocessed, wherein the preprocessing comprises null filling, and/or value mapping.
In a possible implementation, the financial features in the push feature include a financial product feature and a financial product cross feature, wherein the financial product cross feature is determined based on data generated by a financial product cross with a user and/or advertising scene.
Referring to fig. 2, a first embodiment of the advertisement pushing method of the present application includes:
step S10, extracting pushing characteristics from a pushing scene;
the push scenario refers to pushing a personalized advertisement to a user. I.e. advertisements pushed against different users should also be different when the advertisements are pushed. For example, when determining whether to push the advertisement a to the user a, the relevant features of the user a and the relevant features of the advertisement a need to be extracted as push features. The pushing features can be historical consumption records, consumption statistics features and financial statistics features of the user, advertisement content, advertisement expression forms and the like, and product features and the like.
For example, the push feature may be extracted from data generated in a short period of time, for example, the user related total feature is easy to change, so that the total value in the sliding window closest to the current moment may be taken as the user related push feature, for example, generated from data of three months or one month.
Step S20, based on the pushing characteristics, predicting the purchase probability of the pushed product in the pushing scene and the click probability of the recommended advertisement of the pushed product;
The method includes the steps that after the push feature is extracted, the push feature can be input into two prediction models, the two prediction models are respectively predicted, and after a push object pushes a recommended advertisement in a push scene, the click probability that the push object clicks the recommended advertisement and the purchase probability that the push object purchases a pushed product in the recommended advertisement are obtained. Based on the above example, if the pushing scenario a is that an advertisement a is recommended to the user a, and the advertisement a includes a product a, the pushing features a extracted from the pushing scenario are respectively input into two different prediction models, so as to predict and obtain the click probability that the user a clicks the recommended advertisement a, and the purchase probability that the user a purchases the product a.
In a possible implementation manner, the pushing features include a user feature, an advertisement feature, an interaction feature and a financial feature, and the step of predicting the purchase probability of the pushed product in the pushing scene and the click probability of the recommended advertisement of the pushed product based on the pushing features includes:
step S210, inputting the user characteristics, the advertisement characteristics and the interaction characteristics into a first prediction model for prediction to obtain click probability of the recommended advertisement;
and step S220, inputting the user characteristics, the advertisement characteristics, the interaction characteristics and the financial characteristics into a second prediction model to obtain the purchase probability of the pushed product.
The push features include user features, advertisement features, interaction features, and financial features. Wherein, the user characteristics can comprise basic information, financial information, consumption information and the like of the user stored in the database, and can be obtained through a related database table; the advertisement features comprise basic features and derived features, wherein the basic features comprise advertisement effectiveness, delivery positions, advertisement types and the like, and the derived features comprise related image features, text features and the like extracted by analyzing advertisement material pictures through an algorithm; the interaction features include user behavior features from user behavior at APP (Application) and behavior sequence features including SVD (Singular Value Decomposition ) features and deep behavior sequence features. And the financial characteristics are characteristics associated with the financial products recommended by the advertisement.
In a possible embodiment, the financial features in the push feature include a financial product feature and a financial product cross feature, wherein the financial product cross feature is determined based on data generated by a financial product cross with a user and/or advertising scene.
Illustratively, the financial features include financial product features and financial product cross-over features. The characteristics of the financial product are the attribute of the financial product (such as product type, product risk level, annual return, etc.), the transaction related characteristics of the financial product (such as statistic characteristics of sales of the financial product, etc.), etc. And financial product intersection characteristics are determined based on data generated by the financial product intersecting the user and/or advertising scene. Such as advertisement scene-financial product cross-over features, user-financial product cross-over features, and user-advertisement scene-financial product cross-over features. Wherein, the advertisement scene-financial product cross characteristic comprises purchase statistics characteristic generated by the advertisement material related financial product through the current advertisement scene, and the like. The cross characteristics of the user and the financial products comprise interest preferences of the user on the financial products related to the advertisement materials, and the like, and are mainly represented by statistics of browsing behaviors of the user on the related financial products, warehouse holding of the user on the related financial products, and the like. The user-advertising scene-financial product intersection feature is to include a user's purchase statistics through the current advertisement, etc.
And inputting the user characteristics, the advertisement characteristics and the interaction characteristics into a first prediction model for prediction to obtain the click probability of the recommended advertisement. And inputting the user characteristics, the advertisement characteristics, the interaction characteristics and the financial characteristics into a second prediction model to obtain the purchase probability of the pushed product. The first prediction model and the second prediction model may be respectively obtained by performing classification training through a LightGBM model. It should be noted that, in general, the output of the classification model is a classification result, for example, the user clicks on an advertisement and the user does not click on an advertisement, or the user purchases a product or does not purchase a product. Further, taking clicking and non-clicking as examples, the first prediction model may obtain clicking probability and non-clicking probability, if the clicking probability is greater than the non-clicking probability, the classification result of the first prediction model should be clicking, and in this embodiment, the clicking probability is taken as the output result of the first prediction model, and similarly, the purchase probability is taken as the output result of the second prediction model.
Step S30, based on the purchase probability and the click probability, predicting a push value of pushing the recommended advertisement to a push object in the push scene;
illustratively, a push value for pushing the recommended advertisement to a push object in the push scenario is predicted based on the purchase probability and the click probability. Based on the above example, the advertisement pushing system pushes the pushing value of advertisement a to user a, where the pushing value is used to indicate the pushing tendency, and in general, the higher the pushing value, the higher the tendency of advertisement a to be pushed to user a.
In a possible implementation manner, the step of predicting the push value of pushing the recommended advertisement to the push object in the push scene based on the purchase probability and the click probability includes:
step S310, inputting the purchase probability and the click probability into a preset fusion task model to obtain a pushing value for pushing the recommended advertisement to the pushing object, wherein a first weight of the purchase probability and a second weight of the click probability in the preset fusion task model are generated through an adaptive weight generation network.
The purchase probability and the click probability are input into a preset fusion task model, and the preset fusion task model predicts a pushing value for pushing the recommended advertisement to the pushing object. The preset fusion task model can be a classification model. It should be noted that, the first weight of the purchase probability and the second weight of the click probability in the preset fusion task model are generated through the adaptive weight generating network. The preset fusion task model can balance the relation and the emphasis point between each task according to the pre-defined proportional weight in the training process, and the self-adaptive searching of the optimal solution (namely, the first weight and the second weight are automatically adjusted) can be realized through the self-adaptive weight generation network only by adjusting parameters. The preset fusion task model can efficiently integrate the prediction results of the first prediction model and the second prediction model for recommendation, so that the final performance of the whole pushing method is improved, namely the return rate of the advertisement after being pushed is improved, namely the click probability and the purchase probability are synchronously improved. The embodiment utilizes the advantages (click probability prediction and purchase probability prediction) of different tasks, and dynamically adjusts the weights between the tasks in the training process, so that the effect of synchronously improving the click probability and the purchase probability is achieved.
And step S40, pushing recommended advertisements to the pushing objects based on the pushing values.
For example, the push value is a basis for whether to push the push value corresponding to the recommended advertisement to the recommended object, for example, if there are multiple recommended advertisements, the push value of each recommended advertisement may be ranked, and the recommended advertisement with the top ranking may be pushed to the recommended object (for example, ranking in order of the push value from big to small, and pushing the recommended advertisement with the top preset ranking to the recommended object, for example, pushing the recommended advertisement with the top three recommended advertisements to the recommended object). Or comparing the push value with a preset push threshold, and if the push value is greater than the preset push threshold, pushing the recommended advertisement to the push object.
Referring to fig. 3, an overall model framework diagram for applying an advertisement push method is shown. In the figure, the extracted user features, advertisement features and interaction features are input into a first prediction model to obtain click probability, the extracted user features, advertisement features, interaction features and financial features are input into a second prediction model to obtain purchase probability, and the click probability and the purchase probability are input into a preset fusion task model. The output result of the preset fusion task model is used as the basis for pushing advertisements to users.
In this embodiment, the push feature is extracted from the push scene; based on the pushing characteristics, predicting the purchase probability of the pushed product in the pushing scene and the click probability of the recommended advertisement of the pushed product; based on the purchase probability and the click probability, predicting a push value for pushing the recommended advertisement to a push object in the push scene; and pushing recommended advertisements to the pushing objects based on the pushing values. In other words, when pushing advertisements, the click probability of recommended advertisements and the purchase probability of the pushed products are respectively obtained through the pushing features extracted from the pushing scene, and then the click probability and the purchase probability are integrated to obtain the pushing value of pushing the recommended advertisements to the user, if the pushing value is larger, the recommended advertisements can be pushed to the user. Compared with the pushing method which only takes the click rate as a target and forcibly increases the exposure proportion of the related materials of the financial products in the traditional scheme, the embodiment pushes the advertisement to the user by combining the click probability and the purchase probability, thereby synchronously improving the click rate of the advertisement and the purchase rate of the products on the advertisement, namely improving the pushing return rate of the product advertisements.
Referring to fig. 4, a second embodiment of the present application is proposed based on the first embodiment of the present application, and the same parts as those of the above embodiment in this embodiment can be referred to the above, and will not be repeated here. Before the step of predicting a purchase probability of the pushed product and a click probability of the recommended advertisement of the pushed product in the push scenario based on the push feature, the method comprises:
Step A10, performing iterative training on an initialization model based on a preset training sample set to obtain a first prediction model and a second prediction model, wherein the types of training samples in the preset training sample set comprise a first training sample and a second training sample, the labels of the first training sample are clicked or not clicked, the labels of the second training sample are purchased or not purchased, the first prediction model is obtained by training the first training sample, and the second prediction model is obtained by training the second training sample.
For example, in the present embodiment, the first prediction model and the second prediction model may be trained based on the same initialization model. For example, the initialization model may be a LightGBM model. And training by using different training samples. For example, the categories of training samples in the preset training sample set include a first training sample and a second training sample. Wherein the first training sample may be a sample for click probability. And training the initialization model based on the first training sample to obtain a first prediction model. The second training sample may be a sample for purchase probability. And training the initialization model based on the second training sample to obtain a second prediction model. It should be noted that, the types of the labels of the first training sample and the second training sample are different, and the label of the first training sample is clicked or not clicked, and the label of the second training sample is purchased or not purchased, for example, different types of labels may be marked on the same sample to obtain the first training sample and the second training sample respectively, for example, the clicked and not clicked are the same type of label, and the clicked and purchased, clicked and not purchased, not clicked and not purchased are different types of labels. In addition, in the case where the first prediction model and the second prediction model are predicted, the input features are also different. For example, the input of the first predictive model may include user features, advertisement features, and interaction features. While the input of the second predictive model may include user features, advertisement features, interaction features, and financial features. When the method is used for inputting, the user characteristics, the advertisement characteristics and the interaction characteristics can be spliced into the wide table characteristics input by the first prediction model; user features, advertisement features, interaction features, and financial features may be stitched into a broad-table feature input by the second predictive model.
In a possible implementation manner, the initialization model includes a first initialization model and a second initialization model, and the step of iteratively training the initialization model based on a preset training sample set to obtain the first prediction model and the second prediction model includes:
step A110, for any first training sample, inputting the first training sample into the first initialization model to obtain a first prediction result of the first initialization model;
step a120, updating parameters in the first initialization model based on the difference between the first prediction result and the label of the first training sample, and returning to execute the step of inputting the first training sample into the first initialization model based on a new first training sample to obtain a first prediction result of the first initialization model, until a preset training completion condition is reached, and taking the first initialization model as the first prediction model;
step A130, for any second training sample, inputting the second training sample into the second initialization model to obtain a second prediction result of the second initialization model;
and step A140, updating parameters in the second initialization model based on the difference between the second prediction result and the label of the second training sample, and returning to execute the step of inputting the second training sample into the second initialization model based on a new second training sample to obtain a second prediction result of the second initialization model until a preset training completion condition is reached, wherein the second initialization model is used as the second prediction model.
It should be noted that, the initialization model includes a first initialization model and a second initialization model, and the model base structures of the first initialization model and the second initialization model may be the same or different.
For any one first training sample, the first training sample is input into a first initialization model, so that a first prediction result of the first initialization model can be obtained. And comparing the first prediction result with the label of the first training sample, calculating the prediction loss of the first initialization model based on the difference between the first prediction result and the label of the first training sample, and updating model parameters in the first initialization model based on the prediction loss. And repeating the training process based on the new first training sample until the training is completed, and taking the first initialization model at the moment as a first prediction model. The preset training completion condition may be set by a technician with more demands, for example, reaching a preset training number, or model prediction loss converging, etc.
Similarly, for the training process for obtaining the second prediction model, reference may be made to the above process, where only the training samples and the features input by the model are different, and the specific training process will not be described herein.
In a possible embodiment, before the step of inputting the first training sample into the first predictive model and/or before the step of inputting the second training sample into the second predictive model, the method comprises:
step a101, preprocessing the first training sample, and/or preprocessing the second training sample, where the preprocessing includes null filling, and/or value mapping.
Illustratively, the first training sample and/or the second training sample may be preprocessed before the first training sample is input to the first prediction model and/or before the second training sample is input to the second prediction model during a training phase, where the content of the preprocessing includes null padding and/or value mapping, such as padding missing features in the first training sample or the second training sample based on preset padding features (e.g., feature 0), and/or mapping class-type features in the first training sample or the second training sample to feature values, such as by LabelEncoder, while other features may directly use the original values. In addition, it should be noted that, in addition to preprocessing training samples during the training phase, preprocessing push features may also be performed during the application phase.
Referring to fig. 5, in addition, an embodiment of the present application further provides an advertisement pushing device 100, where the advertisement pushing device 100 includes:
an extraction module 10, configured to extract a push feature from a push scene;
a first prediction module 20, configured to predict, based on the push feature, a purchase probability of a pushed product in the push scene and a click probability of a recommended advertisement of the pushed product;
a second predicting module 30, configured to predict a push value of pushing the recommended advertisement to a push object in the push scene based on the purchase probability and the click probability;
a pushing module 40, configured to push a recommended advertisement to the push object based on the push value.
Optionally, the push features include a user feature, an advertisement feature, an interaction feature, and a financial feature, and the first prediction module 20 is further configured to:
inputting the user characteristics, the advertisement characteristics and the interaction characteristics into a first prediction model for prediction to obtain click probability of the recommended advertisement;
and inputting the user characteristics, the advertisement characteristics, the interaction characteristics and the financial characteristics into a second prediction model to obtain the purchase probability of the pushed product.
Optionally, the second prediction module 30 is further configured to:
and inputting the purchase probability and the click probability into a preset fusion task model to obtain a pushing value for pushing the recommended advertisement to the pushing object, wherein a first weight of the purchase probability and a second weight of the click probability in the preset fusion task model are generated through a self-adaptive weight generation network.
Optionally, the advertisement pushing device 100 further includes a training module 50, where the training module 50 is configured to:
the method comprises the steps of carrying out iterative training on an initialization model based on a preset training sample set to obtain a first prediction model and a second prediction model, wherein the types of training samples in the preset training sample set comprise a first training sample and a second training sample, the labels of the first training sample are clicked or not clicked, the labels of the second training sample are purchased or not purchased, the first prediction model is obtained through training of the first training sample, and the second prediction model is obtained through training of the second training sample.
Optionally, the initialization model includes a first initialization model and a second initialization model, and the training module 50 is further configured to:
for any first training sample, inputting the first training sample into the first initialization model to obtain a first prediction result of the first initialization model;
Updating parameters in the first initialization model based on the difference between the first prediction result and the label of the first training sample, returning to execute the step of inputting the first training sample into the first initialization model based on the new first training sample to obtain a first prediction result of the first initialization model until a preset training completion condition is reached, and taking the first initialization model as the first prediction model;
for any second training sample, inputting the second training sample into the second initialization model to obtain a second prediction result of the second initialization model;
updating parameters in the second initialization model based on the difference between the second prediction result and the label of the second training sample, and returning to execute the step of inputting the second training sample into the second initialization model based on the new second training sample to obtain a second prediction result of the second initialization model until a preset training completion condition is reached, wherein the second initialization model is used as the second prediction model.
Optionally, the advertisement pushing device 100 further includes a preprocessing module 60, where the preprocessing module 60 is configured to:
The first training sample is preprocessed, and/or the second training sample is preprocessed, wherein the preprocessing comprises null filling, and/or value mapping.
Optionally, the financial features in the push feature include a financial product feature and a financial product cross feature, wherein the financial product cross feature is determined based on data generated by a financial product cross with a user and/or advertising scene.
The advertisement pushing device provided by the application adopts the advertisement pushing method in the embodiment, and aims to solve the technical problem of low advertisement pushing return rate of the current financial products. Compared with the prior art, the beneficial effects of the advertisement pushing device provided by the embodiment of the application are the same as those of the advertisement pushing method provided by the embodiment, and other technical features of the advertisement pushing device are the same as those disclosed by the method of the embodiment, so that the description is omitted herein.
In addition, to achieve the above object, the present application also provides an electronic device including: the advertisement pushing system comprises a memory, a processor and an advertisement pushing program which is stored in the memory and can run on the processor, wherein the advertisement pushing program realizes the steps of the advertisement pushing method when being executed by the processor.
The specific implementation manner of the electronic device of the present application is substantially the same as the above embodiments of the advertisement pushing method, and will not be described herein.
In addition, in order to achieve the above object, the present application further provides a storage medium having stored thereon an advertisement pushing program which, when executed by a processor, implements the steps of the advertisement pushing method as described above.
The specific implementation manner of the storage medium of the present application is basically the same as that of each embodiment of the advertisement pushing method, and will not be repeated here.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The foregoing embodiment numbers of the present application are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) as described above, comprising instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method according to the embodiments of the present application.
The foregoing description is only of the preferred embodiments of the present application, and is not intended to limit the scope of the application, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.

Claims (10)

1. An advertisement pushing method is characterized by comprising the following steps:
Extracting pushing characteristics from a pushing scene;
based on the pushing characteristics, predicting the purchase probability of the pushed product in the pushing scene and the click probability of the recommended advertisement of the pushed product;
based on the purchase probability and the click probability, predicting a push value for pushing the recommended advertisement to a push object in the push scene;
and pushing recommended advertisements to the pushing objects based on the pushing values.
2. The advertisement pushing method of claim 1, wherein the pushing features include user features, advertisement features, interaction features, and financial features, and wherein predicting the purchase probability of the pushed product in the pushed scene and the click probability of the recommended advertisement of the pushed product based on the pushing features comprises:
inputting the user characteristics, the advertisement characteristics and the interaction characteristics into a first prediction model for prediction to obtain click probability of the recommended advertisement;
and inputting the user characteristics, the advertisement characteristics, the interaction characteristics and the financial characteristics into a second prediction model to obtain the purchase probability of the pushed product.
3. The advertisement pushing method of claim 1, wherein the step of predicting a push value for pushing the recommended advertisement to a push object in the push scene based on the purchase probability and the click probability comprises:
And inputting the purchase probability and the click probability into a preset fusion task model to obtain a pushing value for pushing the recommended advertisement to the pushing object, wherein a first weight of the purchase probability and a second weight of the click probability in the preset fusion task model are generated through a self-adaptive weight generation network.
4. The advertisement pushing method according to claim 2, wherein before the step of predicting a purchase probability of a pushed product in the pushed scene and a click probability of a recommended advertisement of the pushed product based on the push feature, the method comprises:
the method comprises the steps of carrying out iterative training on an initialization model based on a preset training sample set to obtain a first prediction model and a second prediction model, wherein the types of training samples in the preset training sample set comprise a first training sample and a second training sample, the labels of the first training sample are clicked or not clicked, the labels of the second training sample are purchased or not purchased, the first prediction model is obtained through training of the first training sample, and the second prediction model is obtained through training of the second training sample.
5. The advertisement pushing method of claim 4, wherein the initialization model includes a first initialization model and a second initialization model, and the step of iteratively training the initialization model based on a preset training sample set to obtain the first prediction model and the second prediction model includes:
For any first training sample, inputting the first training sample into the first initialization model to obtain a first prediction result of the first initialization model;
updating parameters in the first initialization model based on the difference between the first prediction result and the label of the first training sample, returning to execute the step of inputting the first training sample into the first initialization model based on the new first training sample to obtain a first prediction result of the first initialization model until a preset training completion condition is reached, and taking the first initialization model as the first prediction model;
for any second training sample, inputting the second training sample into the second initialization model to obtain a second prediction result of the second initialization model;
updating parameters in the second initialization model based on the difference between the second prediction result and the label of the second training sample, and returning to execute the step of inputting the second training sample into the second initialization model based on the new second training sample to obtain a second prediction result of the second initialization model until a preset training completion condition is reached, wherein the second initialization model is used as the second prediction model.
6. The advertisement pushing method of claim 5, wherein before the step of inputting the first training sample to the first initialization model and/or before the step of inputting the second training sample to the second initialization model, the method comprises:
the first training sample is preprocessed, and/or the second training sample is preprocessed, wherein the preprocessing comprises null filling, and/or value mapping.
7. The advertisement pushing method according to claims 1 to 6, wherein the financial characteristics in the pushing characteristics include financial product characteristics and financial product crossing characteristics, wherein the financial product crossing characteristics are determined based on data generated by crossing a financial product with a user and/or an advertisement scene.
8. An advertisement pushing device, characterized in that the advertisement pushing device comprises:
the extraction module is used for extracting pushing characteristics from the pushing scene;
the first prediction module is used for predicting the purchase probability of the pushed product in the pushed scene and the click probability of the recommended advertisement of the pushed product based on the push characteristics;
The second prediction module is used for predicting a push value of pushing the recommended advertisement to a push object in the push scene based on the purchase probability and the click probability;
and the pushing module is used for pushing the recommended advertisement to the pushing object based on the pushing value.
9. An electronic device comprising a memory, a processor, and an advertisement push program stored on the memory and executable on the processor, wherein: the advertisement pushing program when executed by the processor implements the steps of the advertisement pushing method according to any of claims 1 to 7.
10. A storage medium having stored thereon an advertisement pushing program which when executed by a processor implements the steps of the advertisement pushing method according to any of claims 1 to 7.
CN202311085864.3A 2023-08-25 2023-08-25 Advertisement pushing method and device, electronic equipment and storage medium Pending CN117132326A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311085864.3A CN117132326A (en) 2023-08-25 2023-08-25 Advertisement pushing method and device, electronic equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311085864.3A CN117132326A (en) 2023-08-25 2023-08-25 Advertisement pushing method and device, electronic equipment and storage medium

Publications (1)

Publication Number Publication Date
CN117132326A true CN117132326A (en) 2023-11-28

Family

ID=88850271

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311085864.3A Pending CN117132326A (en) 2023-08-25 2023-08-25 Advertisement pushing method and device, electronic equipment and storage medium

Country Status (1)

Country Link
CN (1) CN117132326A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114663155A (en) * 2022-04-01 2022-06-24 广州华多网络科技有限公司 Advertisement putting and selecting method and device, equipment, medium and product thereof

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114663155A (en) * 2022-04-01 2022-06-24 广州华多网络科技有限公司 Advertisement putting and selecting method and device, equipment, medium and product thereof

Similar Documents

Publication Publication Date Title
CN111079022B (en) Personalized recommendation method, device, equipment and medium based on federal learning
CN109831684B (en) Video optimization recommendation method and device and readable storage medium
CN108805594B (en) Information pushing method and device
KR20200123015A (en) Information recommendation method, apparatus, device and medium
CN111506849A (en) Page generation method and device
CN109791562B (en) Improving post-installation application interactions
CN112115363A (en) Recommendation method, computing device and storage medium
CN110827112B (en) Deep learning commodity recommendation method and device, computer equipment and storage medium
CN109922379B (en) Advertisement video optimization method, device and equipment and computer readable storage medium
CN110929021A (en) Text information generating method and text information generating device
WO2020221022A1 (en) Service object recommendation method
CN117132326A (en) Advertisement pushing method and device, electronic equipment and storage medium
CN110598120A (en) Behavior data based financing recommendation method, device and equipment
CN104572775B (en) Advertisement classification method, device and server
KR20190007669A (en) Method for providing service of personalized recommendation based on e-mail and apparatus therefor
CN115147130A (en) Problem prediction method, apparatus, storage medium, and program product
CN111310035A (en) Recommendation method and device based on psychological and behavioral characteristics
CN111914173B (en) Content processing method, device, computer system and storage medium
CN114265777A (en) Application program testing method and device, electronic equipment and storage medium
CN113554448A (en) User loss prediction method and device and electronic equipment
CN110837596A (en) Intelligent recommendation method and device, computer equipment and storage medium
CN111915339A (en) Data processing method, device and equipment
CN111984839A (en) Method and apparatus for rendering a user representation
CN113743636B (en) Target operation prediction method and device, electronic equipment and storage medium
US11907508B1 (en) Content analytics as part of content creation

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