CN115601088A - Object delivery method and device, storage medium and electronic equipment - Google Patents

Object delivery method and device, storage medium and electronic equipment Download PDF

Info

Publication number
CN115601088A
CN115601088A CN202211314115.9A CN202211314115A CN115601088A CN 115601088 A CN115601088 A CN 115601088A CN 202211314115 A CN202211314115 A CN 202211314115A CN 115601088 A CN115601088 A CN 115601088A
Authority
CN
China
Prior art keywords
target
user
target user
objects
preset
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
CN202211314115.9A
Other languages
Chinese (zh)
Inventor
费婷婷
陈鸿翔
周庭庭
罗川江
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.)
Hangzhou Netease Cloud Music Technology Co Ltd
Original Assignee
Hangzhou Netease Cloud Music Technology 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 Hangzhou Netease Cloud Music Technology Co Ltd filed Critical Hangzhou Netease Cloud Music Technology Co Ltd
Priority to CN202211314115.9A priority Critical patent/CN115601088A/en
Publication of CN115601088A publication Critical patent/CN115601088A/en
Pending legal-status Critical Current

Links

Images

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
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • 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
    • G06Q30/0255—Targeted advertisements based on user history

Abstract

The disclosure provides an object delivery method, an object delivery device, a storage medium and electronic equipment, and relates to the technical field of computers. According to the method, firstly, responding to an object launching request of a target user on a target page, determining a plurality of hit objects corresponding to a launching condition of a configuration label according to the configuration label corresponding to the target user and a launching condition configured in advance on the target page, calculating a prediction score corresponding to the hit objects by using a pre-trained object recommendation model, determining a plurality of objects to be launched matched with the target user, determining a target launching object corresponding to the target user from the objects to be launched according to a preset launching strategy, and pushing the target launching object to the target user on the target page in a preset launching mode. Therefore, the object to be launched is determined through the configuration tag of the user and the launching condition on the page, so that the selected object to be launched better accords with the preference of the target user, and the hit rate of the user for clicking the launched object can be improved to a certain extent.

Description

Object delivery method and device, storage medium and electronic equipment
Technical Field
The embodiment of the disclosure relates to the technical field of computers, and more particularly, to an object delivery method, an object delivery device, a storage medium and an electronic device.
Background
In content distribution, "recommendations" are the most prevalent, and are also the best known way of distribution. The recommendation system is a set of system which tries to identify the user interest by using data such as user basic attributes and user historical behaviors, screens out the content which is most matched with the user portrait from the whole content pool, recommends the content to the user, collects feedback data of the user on the recommended content, continuously iterates the user portrait and recommends the user interest based on the user portrait.
Although the recommendation system can match the user information and the content information with each other for efficient personalized recommendation, there are many disadvantages that are not negligible. On the one hand, sometimes user behavior is not equivalent to user interests. For example, some content of the title party and the edge ball are disliked, but the user often cannot endure clicking when seeing the content, and clicking, watching and other actions are generally regarded as positive feedback actions in the recommendation system, and the recommendation system regards that the user likes the content, so that the recommendation of the content is increased, a vicious circle is formed, and the user experience is hurt. On the other hand, the "malay effect" problem exists, on the content side, in millions of levels of content pools, 10% of content at the head possibly occupies 80% of flow, so that the content distribution at the middle waist part is insufficient, the high-quality content at the long tail cannot be effectively exposed, the situation of constant intensity of a strong person is formed, and the development of the long-term ecology of the platform content is not facilitated.
Disclosure of Invention
In order to overcome the problems in the related art, the present disclosure provides an object delivery method, an object delivery apparatus, a storage medium, and an electronic device.
According to a first aspect of the present disclosure, there is provided an object delivery method, including:
responding to an object launching request of a target user on a target page, and determining a plurality of hit objects corresponding to launching conditions meeting the configuration tags according to the configuration tags corresponding to the target user and launching conditions configured in advance on the target page;
calculating the prediction scores corresponding to the hit objects by using a pre-trained object recommendation model to determine a plurality of objects to be delivered matched with the target user; the object recommendation model is a model for determining the target user's preference for the hit object, the target user's characteristics, and the hit object's heat;
determining a target release object corresponding to the target user from the plurality of objects to be released according to a preset release strategy;
and pushing the target delivery object to the target user on the target page in a preset delivery mode.
According to a second aspect of the present disclosure, there is provided an object delivery apparatus, the apparatus comprising:
the system comprises a first determining module, a second determining module and a third determining module, wherein the first determining module is used for responding to an object launching request of a target user on a target page, and determining a plurality of hit objects corresponding to launching conditions of configuration tags according to the configuration tags corresponding to the target user and launching conditions configured in advance on the target page;
the calculation module is used for calculating the prediction scores corresponding to the hit objects by using a pre-trained object recommendation model to determine a plurality of objects to be delivered matched with the target user; the object recommendation model is a model for determining the target user's preference for the hit object, the target user's characteristics, and the hit object's heat;
the second determining module is used for determining a target release object corresponding to the target user from the plurality of objects to be released according to a preset release strategy;
and the pushing module is used for pushing the target delivery object to the target user on the target page in a preset delivery mode.
According to an aspect of the present disclosure, there is provided a storage medium having stored thereon a computer program which, when executed by a processor, performs the object delivery method described above.
According to an aspect of the present disclosure, there is provided an electronic device including:
a processor; and
a memory for storing executable instructions of the processor;
wherein the processor is configured to perform any one of the object delivery methods described above via execution of the executable instructions.
To sum up, the object launching method provided by the embodiment of the present disclosure may first respond to an object launching request of a target user on a target page, determine a plurality of hit objects corresponding to launching conditions of configuration tags according to configuration tags corresponding to the target user and launching conditions pre-configured on the target page, calculate predicted scores corresponding to the plurality of hit objects by using a pre-trained object recommendation model, and determine a plurality of objects to be launched matched by the target user, where the object recommendation model is a model for determining preferences of the target user on the hit objects, characteristics of the target user, and heat of the hit objects, determine a target launching object corresponding to the target user from the plurality of objects to be launched according to a preset launching policy, and push the target launching object to the target user on the target page in a preset launching manner. Therefore, the objects to be released are determined through the configuration tags of the users and the releasing conditions on the pages, so that the selected objects to be released are more in line with the preference of the target users, the hit rate of the users for clicking the released objects can be improved to a certain extent, the target releasing objects corresponding to the target users are selected based on the preset releasing strategy, the Martian effect of flow distribution during object recommendation can be reduced, more different target releasing objects are pushed to the target users on the basis that the releasing objects are in line with the preference of the target users, and the releasing efficiency of the released objects can be improved.
Drawings
The above and other objects, features and advantages of exemplary embodiments of the present disclosure will become readily apparent from the following detailed description read in conjunction with the accompanying drawings. Several embodiments of the present disclosure are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which:
fig. 1 is a flowchart illustrating steps of an object delivery method according to an embodiment of the present disclosure;
FIG. 2 is a flow chart of determining a hit object provided by an embodiment of the present disclosure;
FIG. 3 is a schematic diagram for determining a hit object according to an embodiment of the present disclosure;
fig. 4 is a flowchart of determining a target delivery crowd according to an embodiment of the present disclosure;
fig. 5 is a flowchart for determining an object to be delivered according to an embodiment of the present disclosure;
FIG. 6 is a schematic diagram illustrating a calculation of hit prediction scores provided in implementations of the present disclosure;
fig. 7 is a flowchart of determining a target delivery object according to an embodiment of the present disclosure;
fig. 8 is a block diagram of a subject delivery apparatus according to an embodiment of the present disclosure;
FIG. 9 is a schematic illustration of a storage medium according to an embodiment of the present disclosure; and
fig. 10 is a block diagram of an electronic device according to an embodiment of the disclosure.
In the drawings, the same or corresponding reference numerals indicate the same or corresponding parts.
Detailed Description
The principles and spirit of the present disclosure will be described below with reference to several exemplary embodiments. It is understood that these embodiments are given solely for the purpose of enabling those skilled in the art to better understand and to practice the present disclosure, and are not intended to limit the scope of the present disclosure in any way. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
As will be appreciated by one of skill in the art, embodiments of the present disclosure may be embodied as a system, apparatus, device, method, or computer program product. Accordingly, the present disclosure may be embodied in the form of: entirely hardware, entirely software (including firmware, resident software, micro-code, etc.), or a combination of hardware and software, the data involved in the present disclosure may be data that is authorized by the user or fully authorized by various parties. The collection, transmission, use, etc. of the data all meet the requirements of relevant national laws and regulations, and the embodiments/examples of the disclosure can be combined with each other.
In this document, any number of elements in the drawings is intended to be illustrative and not restrictive, and any nomenclature is used for distinction only and not for any restrictive meaning.
The principles and spirit of the present disclosure are explained in detail below with reference to several representative embodiments of the present disclosure.
Fig. 1 is a flowchart of steps of a method for delivering an object according to an embodiment of the present disclosure, and as shown in fig. 1, the method may include:
step S101, responding to an object delivery request of a target user on a target page, and determining a plurality of hit objects corresponding to delivery conditions according to a configuration tag corresponding to the target user and delivery conditions configured in advance on the target page.
In the embodiment of the present disclosure, the object delivery request of the target user on the target page may be an object delivery request triggered when the target user enters the target page, and since an interface recommending a delivery object to the user may be provided on the page, and the interface is used to play the delivery object to the user when the user enters the target page, when the target user enters the target page, the interface is equivalent to an interface which starts the recommended delivery object on the target page, that is, when the target user enters the target page, the object delivery request of the interface for the recommended delivery object is equivalent to an object delivery request which is sent. The configuration tag corresponding to the target user may be a configuration tag added in advance according to the feature information set by the user, and specifically, the configuration tag may be determined by the user feature information of multiple dimensions, such as gender, age, location, occupation, hobby, and the like. The delivery conditions preconfigured on the target page may be delivery conditions preconfigured for different pages, that is, different delivery conditions may be configured in a personalized manner according to the content displayed on each page, so as to improve the purpose of the user clicking the delivery object, where the delivery conditions may include plan configuration, anchor configuration, and crowd configuration, where the plan configuration may be a configuration that a plan delivery time period, a delivery policy, a delivery anchor, and a delivery crowd need to be configured, the delivery anchors in the plan configuration may be associated with the anchor configuration synchronously, and the details of the anchor configuration may include: the method comprises the following steps of anchor ID, anchor type, anchor delivery time period, anchor delivery flow upper limit and lower limit threshold. The crowd configuration can mainly be to configure crowd labels, the crowd labels support basic attribute labels such as 'age', 'gender', 'city', and the like, live broadcast payment labels such as 'live broadcast new user', 'live broadcast active user', 'live broadcast payment user', and the like, and other user interest labels obtained from the behavior of the user in the application.
In the embodiment of the present disclosure, in response to an object placement request of a target user on a target page, according to a configuration tag corresponding to the target user and a placement condition preconfigured on the target page, a plurality of hit objects corresponding to a placement condition that meets the configuration tag are determined, which may be that after the target user enters the target page, in response to the object placement request of the target user on the target page, the configuration tag corresponding to the target user and a placement condition preconfigured on the target page are determined, the placement condition that meets the configuration tag is determined according to the configuration tag and the placement condition, and finally, a plurality of hit objects corresponding to the placement condition that meets the configuration tag are determined. Wherein, each releasing condition corresponds to a plurality of objects for releasing to users, and the objects can be anchor, commodities, videos, links and the like.
Step S102, calculating the corresponding prediction scores of the hit objects by using a pre-trained object recommendation model to determine a plurality of objects to be delivered matched with the target user; the object recommendation model is a model for determining the target user's preferences for the hit object, the target user's characteristics, and the hit object's heat.
In the embodiment of the present disclosure, the object recommendation model may be obtained by pre-training an initial object recommendation model by using sample features. Specifically, the object recommendation model may be iteratively trained by continuously using the sample features, so that the object recommendation model may learn the prediction capability of correctly generating the user preference degree, the user preference prediction capability, and the object popularity prediction capability according to the input features. Therefore, in the embodiment of the present disclosure, the trained object recommendation model may be used to calculate the user characteristics, the content-related characteristics, the content-side characteristics or the user characteristics, or the content-side characteristics, so as to obtain the prediction scores of the hit object preferences, the user preferences, and the hit object popularity represented by each characteristic. The object recommendation model may be a Deep recommendation model (Deep fm), a click-through rate prediction Model (MLR), or another refined model for Deep learning.
In the embodiment of the disclosure, the pre-trained object recommendation model is used to calculate the prediction scores corresponding to the hit objects to determine the multiple objects to be launched matched by the target user, the prediction score of each hit object may be calculated by the pre-trained object recommendation model, and the hit objects with the prediction scores meeting the preset score threshold are used as the objects to be launched, so as to obtain the multiple objects to be launched matched by the target user.
Step S103, determining a target delivery object corresponding to the target user from the plurality of objects to be delivered according to a preset delivery strategy.
In the embodiment of the present disclosure, the preset delivery policy may be a delivery policy preset on a target page, where different delivery policies may be set on different pages. In order to meet diversified operation requirements, the preset releasing strategy can be any one of releasing strategies such as random supporting, weighted supporting, top supporting and frequency control. Specifically, the random support can be simply that random distribution of crowd matching users is carried out on the anchor hit by the delivery plan according to probability; the top support can be to sort and distribute the anchor hit by plan according to the score of the fine-ranking model; the weighting support can be that the anchor hit by the plan is weighted and the weighting factors with adjustable fusion parameters are added for sequencing and distribution on the basis of obtaining the prediction score by calculation; the frequency control may be to limit the number of times the anchor is repeatedly exposed to the same user, prevent user fatigue,
in the embodiment of the disclosure, a target delivery object corresponding to a target user is determined from a plurality of to-be-delivered objects according to a preset delivery policy, which may be that the plurality of to-be-delivered objects are screened based on the preset delivery policy on a target page, and the to-be-delivered objects meeting the preset delivery policy are taken as the target delivery objects recommended to the target user.
And step S104, pushing the target delivery object to the target user on the target page according to a preset delivery mode.
In the embodiment of the disclosure, since the target page may be provided with the region for displaying the launched object, the target launched object may be displayed in the preset region of the target page according to a preset launch mode, so as to push the target launched object to the target user. The preset releasing mode can be a preset releasing mode for displaying released objects, and the preset releasing mode can be releasing modes such as top display and suspension display, and the preset releasing mode is not limited by the disclosure.
To sum up, the object delivery method provided by the embodiment of the present disclosure may first respond to an object delivery request of a target user on a target page, determine, according to a configuration tag corresponding to the target user and a delivery condition pre-configured on the target page, a plurality of hit objects corresponding to the delivery condition that meets the configuration tag, calculate, using a pre-trained object recommendation model, prediction scores corresponding to the plurality of hit objects, and determine a plurality of objects to be delivered that are matched by the target user, where the object recommendation model is a model for determining preferences of the target user for the hit objects, characteristics of the target user, and heat of the hit objects, determine, according to a preset delivery policy, a target delivery object corresponding to the target user from the plurality of objects to be delivered, and push the target delivery object to the target user on the target page in a preset delivery manner. Therefore, the objects to be released are determined through the configuration tags of the users and the releasing conditions on the pages, so that the selected objects to be released are more in line with the preference of the target users, the hit rate of the users for clicking the released objects can be improved to a certain extent, the target releasing objects corresponding to the target users are selected based on the preset releasing strategy, the Matai effect of flow distribution during object recommendation can be reduced, more different target releasing objects are pushed to the target users on the basis that the releasing objects are in line with the preference of the target users, and the releasing efficiency of the released objects can be improved.
Optionally, the pre-configured launch conditions in this embodiment of the present disclosure may include a launch crowd and a corresponding launch object, and the determining, according to the configuration tag corresponding to the target user and the launch conditions pre-configured on the target page, operations of multiple hit objects corresponding to the launch conditions that meet the configuration tag may specifically include, as shown in fig. 2:
step S1011, determining each dimension feature information included in the target user based on the configuration label corresponding to the target user.
In the embodiment of the present disclosure, the configuration tag corresponding to the target user may be a tag selected by the target user by self-service to add, or may be a tag generated according to basic information filled by the user, so that each piece of dimensional feature information included in the target user may be determined according to the dimension represented by the configuration tag corresponding to the target user and the content of the configuration tag. For example, if the target user has configuration tags "male", "35 years old" and "engineer", it may be determined that the target user includes dimension feature information "sex-male", "age-35 years old" and "professional-engineer".
And step S1012, determining a target release crowd according with the dimensional feature information from the release crowd according to the dimensional feature information contained in the target user.
In the embodiment of the present disclosure, since the corresponding release crowds of different dimension characteristic information are different, the matched release crowds can be determined according to each dimension characteristic information included in the target user, and the target release crowd to which the target user belongs is obtained.
Step S1013, based on a preset mapping relationship between the target putting crowd and the putting objects, determining the plurality of putting objects corresponding to the target putting crowd as a plurality of hit objects corresponding to the putting conditions that meet the configuration tag.
In the embodiment of the present disclosure, the preset mapping relationship may be a preset matching mapping relationship between the target release population and the release objects, and therefore, each release object corresponding to the target release population may be determined according to the preset mapping relationship, and the plurality of release objects may be used as a plurality of hit objects corresponding to the release conditions that meet the configuration tag. For example, a recall index may be established by writing a preset mapping relationship between the released population and the released object into an inverted index table, and when the dimensional feature information represented by the user satisfies the user portrait of the released population corresponding to the released object, each released object corresponding to the released population is used as a plurality of hit objects corresponding to the release condition that satisfies the configuration tag.
Optionally, in the embodiment of the present disclosure, the operation of determining the dimensional feature information included in the target user based on the configuration tag corresponding to the target user specifically includes:
determining feature information corresponding to each preset user dimension of the target user according to the configuration label corresponding to the target user; if the feature information of the configuration label corresponding to the preset user dimension is empty information, determining the supplementary information of the target user corresponding to the preset user dimension according to a preset feature filling mode; and obtaining feature information of each dimension contained by the target user according to the feature information and/or the supplementary information of the target user in each preset user dimension.
In the embodiment of the present disclosure, the preset user dimension may be a user dimension preset according to an actual situation, for example, the preset user dimension may be a gender, an age, an occupation, a location, a hobby, and the like, and therefore, the corresponding preset user dimension may be determined according to the configuration tag associated with the target user, and then the corresponding feature information of the target user in each preset user dimension may be determined according to the content of the configuration tag.
In this embodiment of the present disclosure, if there is a certain preset user dimension and the target user does not have a corresponding configuration tag, that is, the feature information corresponding to the configuration tag in the preset user dimension is null information, it may be determined according to a preset feature filling manner 8 And the supplementary information corresponding to the target user in the preset user dimension is determined, wherein the preset feature filling mode can be a feature filling mode preset according to the actual situation. For example, when the configuration tag does not have a specified preset user dimension, that is, the target user may receive the delivery object corresponding to each preset user dimension, so that the supplementary information of the target user in the preset user dimension may be filled in as all; when the configuration tag of the target user lacks information of a certain preset user dimension, the supplementary information of the target user on the preset user dimension can be filled as empty, so that the supplementary information of the preset user dimension is the empty dimension, and only the delivery object with the preset user dimension of all can be matched.
In the embodiment of the present disclosure, the feature information of each dimension included by the target user is obtained according to the feature information and/or the supplementary information of the target user in each preset user dimension, where each dimension feature information is obtained by determining corresponding feature information according to the configuration tag in each preset user dimension, or each dimension feature information is obtained by determining corresponding feature information according to the configuration tag in a part of each preset user dimension and obtaining each dimension feature information according to corresponding supplementary information in another part of each preset user dimension, or each dimension feature information is obtained according to corresponding supplementary information in each preset user dimension. Therefore, after the feature information of each preset user dimension is supplemented by the configuration label, all the preset user dimensions required by the launching condition can be matched during matching, so that a correct launching object corresponding to the configuration label is obtained, and the problem that the number of the matched launching objects is enlarged due to the lack of the feature information of each preset user dimension is solved.
For example, fig. 3 is a schematic diagram for determining a hit object provided by the embodiment of the present disclosure, and as shown in fig. 3, in each preset user dimension on a configuration tag, task1: age =10-30& & gender = all & & region = all, task2: age =10-30& & gender = woman & & region = Shenzhen, a reverse index table is established according to task1 and task2, the reverse index table comprises a reverse key and a reverse value _ list, and when the condition that the input group needs to be screened is age =25& & region = Shenzhen & & gender = empty, finally, the recall result is obtained as task1, and the correct result is obtained as task1, namely, the recall result is the same as the correct result.
It should be noted that, when the user is a new user of the recommendation system, that is, the user does not have a corresponding configuration tag, in order to push a delivery object to the new user, corresponding supplementary information may be added to the feature information of each preset user dimension of the new user according to a preset feature filling manner, that is, the feature information of each dimension included in the new user is obtained according to the supplementary information of the new user in each preset user dimension.
Optionally, in this embodiment of the present disclosure, when a target placement crowd meeting the feature information of each dimension is not found from the placement crowd, as shown in fig. 4, the object placement method may further include:
step S201, determining an extended user similar to the seed user by using a preset similarity calculation method according to the seed user corresponding to the input object; the seed user is a user that matches the user representation associated with the drop object.
In the embodiment of the disclosure, when a target throw-in crowd meeting the feature information of each dimension is not found in the throw-in crowd, a seed user matched with a user portrait associated with a throw-in object is determined, where the seed user may be a preset user matched with the throw-in object one by one, and then an extended user with similarity meeting a threshold between users and the seed user is calculated from a plurality of users by using a preset similarity calculation method. The preset similar calculation method can be a similar population expansion (Lookalike) technology, and the similarity degree of other users and the seed users is calculated Through the preset similar calculation method, so that automatic similar population dynamic expansion is performed, and therefore, on one hand, different requirements of different delivery plans on population orientation accuracy and coverage Rate can be met, on the other hand, the Click Through Rate (Click-Through-Rate, CTR) and the Conversion Rate (CVR) of the expanded population can be obviously improved Through the similar population expansion (Lookalike) technology, the distribution of intra-site flow can be optimized, and the overflow Rate and the commercial change capability of the flow are improved.
It should be noted that, an extended user whose similarity between the user and the seed user meets a threshold is calculated from multiple users by using a preset similarity calculation method, in the embodiment of the present disclosure, a Cosine similarity (Cosine similarity) between user-user vectors (embedding) may be used to measure the similarity between users, where the Cosine similarity may be represented as:
Figure BDA0003908342040000101
wherein, sim cosine (f i ,f j ) May be f i And f j Cosine similarity, f i Can be expressed as a seed user corresponding to the object of delivery, f j Can be represented as other users, f i *f j Can be expressed as the product of the feature vectors between the seed user and the other users, | f i ‖||f j | | may be expressed as a product of the moduli of the feature vectors between the seed user and other users.
And S202, selecting a plurality of users from the extended users and the seed users to form an extended release crowd according with the release plan number according to the release plan number configured by the release object.
In the embodiment of the disclosure, each delivery is used 10 The number of plans to be launched can be pre-configured for the object, therefore, a plurality of users can be selected from the extended users according to the number of the launch plans configured for the launch object, the sum of the number of the selected users and the number of the seed users can be in accordance with the number of the launch plans, and the selected users and the seed users are used as extended launch crowds.
Step S203, determining target throwing population according with the dimension characteristic information from the extended throwing population according to the dimension characteristic information contained in the target user.
In the embodiment of the present disclosure, according to each dimension feature information included in the target user, if the feature information on the dimension of the preset user appears as null information in the target user, the information on the dimension is supplemented in a preset feature filling manner, and then, a target release crowd meeting each dimension feature information is determined from the extended release crowd. Therefore, after the characteristic information on each dimension is supplemented, all directional dimensions required by the task can be screened during screening, and a correct recall result is obtained.
Optionally, the object recommendation model in the embodiment of the present disclosure may include a first recommendation model, a second recommendation model, and a third recommendation model, and the operation of determining the multiple objects to be delivered that are matched by the target user by calculating the prediction scores corresponding to the multiple hit objects by using the pre-trained object recommendation model is specifically included, as shown in fig. 5:
step S1021, based on the configuration tag corresponding to the target user, the delivery condition pre-configured on the target page, and the attribute characteristics included in the hit object, obtaining the user-side characteristics corresponding to the target user, the content-side characteristics corresponding to the hit objects, and the content-related characteristics corresponding to each hit object by the target user.
In the embodiment of the present disclosure, the user-side feature corresponding to the target user may be determined according to the configuration tag corresponding to the target user, the content-side features corresponding to the multiple hit objects are determined according to the attribute features included in the hit objects, and the content-related feature corresponding to each hit object by the target user is determined according to the configuration tag corresponding to the target user, the delivery condition pre-configured on the target page, and the attribute features included in the hit objects. The content-related characteristics may be the content preference, click rate, viewing duration, and other characteristics of the target user for the type of the hit object.
Step S1022, for each hit object, inputting the user-side feature, the content-related feature, and the content-side feature into a pre-trained first recommendation model, and obtaining a first output value representing the preference degree of the target user for the hit object; inputting the user side features into a pre-trained second recommendation model to obtain a second output value representing the user features corresponding to the target user; and inputting the content side features into a pre-trained third recommendation model to obtain a third output value representing the hot value of the hit object.
In the embodiment of the present disclosure, the first recommendation model, the second recommendation model, and the third recommendation model may be a deep recommendation model (deep fm), specifically, the user-side feature, the content-related feature, and the content-side feature may be input to the pre-trained first recommendation model, and an output result of the first recommendation model is used as a first output value representing a preference degree of the target user for the hit object. The user-side features may be input to a pre-trained second recommendation model, and an output result of the second recommendation model is used as a second output value representing the user features corresponding to the target user. The content-side features may be input to a pre-trained third recommendation model, the output of which serves as a third output value characterizing the hot value of the hit object.
Step S1023, regarding each hit object, taking a difference value between a first output value and a bias output value corresponding to the hit object as a prediction score corresponding to the hit object; the offset output value is determined from a product of the second output value and the third output value.
In this embodiment of the present disclosure, a first output value, a second output value, and a third output value calculated for each hit object may be obtained, the first output value, the second output value, and the third output value are input into a preset prediction score formula, an offset output value is determined according to a product of the second output value and the third output value, and then a prediction score corresponding to the hit object is determined according to a difference between the first output value and the offset output value, where the preset prediction score formula may be represented as follows:
y＝y ui -α*y bias ＝y ui -α*y bias_i *y bias_u
where y can be expressed as the predicted score for the hit, y ui Can be expressed as a first output value, y, of the hit bias Can be expressed as the offset output value, y, of the hit bias_i Can be expressed as a second output value, y, of the hit bias_u Can be expressed as a third output value of the hit object, and α can be expressed as a preset empirical parameter, and the value can be between 0 and 1.
Step S1024, selecting the hit object with the prediction score meeting a preset score threshold from the hit objects as the object to be launched matched by the target user.
In the embodiment of the disclosure, for the prediction scores corresponding to the multiple hit objects, the hit objects with the prediction scores meeting the preset score threshold are selected, and all the hit objects with the prediction scores meeting the preset score threshold are used as the objects to be launched matched by the target user, where the preset score threshold may be a value obtained according to actual experience.
Optionally, in the embodiment of the present disclosure, the operation of inputting the user-side feature, the content-related feature, and the content-side feature into a pre-trained first recommendation model to obtain a first output value representing the preference degree of the target user for the hit object may specifically include:
performing vector conversion and splicing on the user side features, the content related features and the content side features to obtain user feature vectors corresponding to the target user; and inputting the user feature vector into the pre-trained first recommendation model to obtain an output result serving as the first output value.
In the embodiment of the disclosure, when the first output value of the hit object is calculated through the first recommendation model, in order to improve efficiency of model calculation, the user-side feature, the content-related feature, and the content-side feature may be converted into vector conversion and vector splicing is performed to obtain a user feature vector corresponding to the target user, the user feature vector is input into the pre-trained first recommendation model for calculation, and an obtained model output result is used as the first output value of the hit object.
Optionally, the object delivery method in the embodiment of the present disclosure may further include:
first, sample user characteristics, sample content-related characteristics, sample content-side characteristics, and true scores for sample objects are obtained.
In the embodiment of the present disclosure, the sample user characteristic, the sample content related characteristic, and the sample content side characteristic may be calculated by a history user, or may be a manually input characteristic, and the real score for the sample object may be a real score representing an actual operation performed on the sample object by the sample user.
And secondly, taking the sample user characteristics, the sample content related characteristics and the sample content side characteristics as first training samples, taking the sample user characteristics as second training samples and taking the sample content side characteristics as third training samples.
In the embodiment of the present disclosure, the sample user characteristics, the sample content-related characteristics, the sample content-side characteristics, and the true score may be used as a first training sample, the sample user characteristics and the true score may be used as a second training sample, and the sample content-side characteristics and the true score may be used as a third training sample.
And finally, performing iterative training on an initial first recommendation model by using the first training sample, performing iterative training on an initial second recommendation model by using the second training sample, and performing iterative training on an initial third recommendation model by using the third training sample to obtain the pre-trained first recommendation model, the pre-trained second recommendation model and the pre-trained third recommendation model of which the output training score accords with the real score.
In the embodiment of the disclosure, a first training sample may be input to an initial first recommendation model for calculation, a second training sample may be input to an initial second recommendation model for calculation, a third training sample may be input to an initial third recommendation model for calculation, a training prediction score may be calculated by a preset prediction score formula, a loss function between the training prediction score and a true score may be calculated, if a result of the loss function does not meet a preset threshold, training parameters in the first recommendation model, the second recommendation model, and the third recommendation model may be adjusted, and an operation of inputting the training sample to the initial recommendation model for calculation may be re-executed until a result of the loss function meets the preset threshold; if the loss function result meets the preset threshold value, that is, it is determined that the training prediction score is approximate to the real score, the first recommendation model obtained through training may be used as a pre-training first recommendation model, the second recommendation model obtained through training may be used as a pre-training second recommendation model, and the second recommendation model obtained through training may be used as a pre-training second recommendation model. Therefore, the model can learn the capability of correctly calculating and processing the sample characteristics to obtain the real prediction score by continuously adopting the sample characteristics for training and adjusting the training parameters of the model.
For example, fig. 6 is a schematic diagram of calculating a hit object prediction score according to an embodiment of the present disclosure, as shown in fig. 6, for each hit object, a user-side feature 301, a content-related feature 302, and a content-side feature 303 of the hit object are determined, the user-side feature 301, the content-related feature 302, and the content-side feature 303 may be subjected to vector conversion and concatenation to obtain a user feature vector 304 corresponding to a target user, the user feature vector 304 is input into a first recommendation model 305, and an output result first output value y is obtained ui 306, inputting the user-side feature 301 into the second recommendation model 307 to obtain a second output value, inputting the content-side feature 303 into the third recommendation model 308 to obtain a third output value, and obtaining the offset output value y by the product of the second output value and the third output value bias 309, and outputting the first output value y ui 306 and an offset output value y bias 309 as the prediction score y310 corresponding to the hit.
Optionally, in the embodiment of the present disclosure, the operation of determining the target delivery object corresponding to the target user from the multiple objects to be delivered according to the preset delivery policy may specifically include, as shown in fig. 7:
and step S1031, determining the exposure time and the exposure times corresponding to each object to be released for the plurality of objects to be released.
In the embodiment of the present disclosure, the exposure time and the exposure times corresponding to the object to be delivered may refer to the time for delivering the object to be delivered to the user and the times for delivering the object to be delivered to the user. For a plurality of objects to be launched, determining the exposure time and the exposure times corresponding to each object to be launched, which may be determining the exposure time and the exposure times corresponding to each object to be launched according to the historical exposure information of each object to be launched for the plurality of objects to be launched.
Step S1032, according to the exposure time and the exposure times corresponding to each object to be launched, selecting an object to be launched, where the exposure time and the exposure times are smaller than a preset exposure threshold, as the target launched object.
In the embodiment of the disclosure, in order to avoid the problem that the drop objects have a martensitic effect, that is, the drop objects with high flow rate obtain more drop times, and the drop objects with low flow rate hardly obtain drop opportunities, the drop objects with exposure time and exposure times smaller than the preset exposure threshold value can be selected as the target drop objects according to the exposure time and exposure times corresponding to each drop object, so that it is ensured that more drop objects can obtain exposure opportunities. For example, a user ID is used as a primary key, a list of anchor IDs exposed to the user and corresponding exposure time stamps are used as key values, and the key values are stored in a database (redis) in a < k, v > key value pair format. When the user requests next time, the user can directly obtain the exposed anchor list from a database (redis) according to the user ID to carry out the exposure and pressing with time + times limitation.
Optionally, in this embodiment of the present disclosure, after the target delivery object is pushed to the target user on the target page in a preset delivery manner, the method may further include:
monitoring feedback information of the target user to the target delivery object; and if the feedback information does not accord with a preset feedback threshold value, adjusting the preset delivery strategy, and reselecting the target delivery object from the plurality of objects to be delivered according to the adjusted delivery strategy.
In the embodiment of the disclosure, in order to ensure the effect of the delivery system, monitoring feedback is particularly important, on one hand, there is an effective evaluation on the content quality of the delivered object, and on the other hand, the influence of the delivery plan on the flow distribution efficiency can also be effectively observed, so that feedback information of a target user on the target delivered object can be monitored, if the feedback information does not meet a preset feedback threshold, a preset delivery strategy is adjusted, and a target delivered object is reselected from a plurality of objects to be delivered according to the adjusted delivery strategy, where the feedback information may be service indexes such as a click rate and a conversion rate of the user on the delivered object. For example, the log-based delivery site performs data analysis and statistics, including supporting large disk data, anchor supporting details, flow statistics of supporting scenes, and the like, and configures the statistical data to a data visualization platform to generate a visualized data billboard.
It should be noted that, in the object delivery method provided in the embodiment of the present disclosure, the execution main body may be an object delivery device, or, a control module in the object delivery device, which is used for executing the loaded object delivery method. In the embodiment of the present disclosure, an object delivery method executed by an object delivery device is taken as an example to describe the object delivery method provided in the embodiment of the present disclosure. Next, an object delivery apparatus according to an exemplary embodiment of the present disclosure will be described with reference to fig. 8.
Fig. 8 schematically shows a block diagram of a subject delivery apparatus according to an embodiment of the present disclosure, and as shown in fig. 8, the subject delivery apparatus 40 may include:
a first determining module 401, configured to, in response to an object placement request of a target user on a target page, determine, according to a configuration tag corresponding to the target user and a placement condition pre-configured on the target page, a plurality of hit objects corresponding to the placement condition that meets the configuration tag;
a calculating module 402, configured to calculate, by using a pre-trained object recommendation model, prediction scores corresponding to the multiple hit objects to determine multiple objects to be delivered that are matched by the target user; the object recommendation model is a model for determining the target user's preference for the hit object, the target user's characteristics, and the hit object's heat;
a second determining module 403, configured to determine, according to a preset delivery policy, a target delivery object corresponding to the target user from the multiple objects to be delivered;
a pushing module 404, configured to push the target delivery object to the target user on the target page in a preset delivery manner.
To sum up, the object delivery device provided in the embodiment of the present disclosure may first respond to an object delivery request of a target user on a target page, determine, according to a configuration tag corresponding to the target user and a delivery condition pre-configured on the target page, a plurality of hit objects corresponding to the delivery condition that meets the configuration tag, calculate, using a pre-trained object recommendation model, prediction scores corresponding to the plurality of hit objects, and determine a plurality of objects to be delivered that are matched by the target user, where the object recommendation model is a model for determining preferences of the target user for the hit objects, characteristics of the target user, and heat of the hit objects, determine, according to a preset delivery policy, a target delivery object corresponding to the target user from the plurality of objects to be delivered, and push the target delivery object to the target user on the target page in a preset delivery manner. Therefore, the objects to be released are determined through the configuration tags of the users and the releasing conditions on the pages, so that the selected objects to be released are more in line with the preference of the target users, the hit rate of the users for clicking the released objects can be improved to a certain extent, the target releasing objects corresponding to the target users are selected based on the preset releasing strategy, the Martian effect of flow distribution during object recommendation can be reduced, more different target releasing objects are pushed to the target users on the basis that the releasing objects are in line with the preference of the target users, and the releasing efficiency of the released objects can be improved.
Optionally, the preconfigured delivery conditions include a delivery crowd and a corresponding delivery object, and the first determining module 401 is further configured to:
determining all dimension characteristic information contained in the target user based on the configuration label corresponding to the target user;
determining a target release crowd according with the dimensional feature information from the release crowd according to the dimensional feature information contained in the target user;
and determining a plurality of throwing objects corresponding to the target throwing crowd as a plurality of hitting objects corresponding to the throwing conditions meeting the configuration label based on a preset mapping relation between the throwing crowd and the throwing objects.
In the embodiment of the present disclosure, the preset mapping relationship may be a preset matching mapping relationship between the target release population and the release objects, and therefore, each release object corresponding to the target release population may be determined according to the preset mapping relationship, and the plurality of release objects may be used as a plurality of hit objects corresponding to the release conditions that meet the configuration tag.
Optionally, the first determining module 401 is further configured to:
determining feature information corresponding to each preset user dimension of the target user according to the configuration label corresponding to the target user;
if the feature information of the configuration label corresponding to the preset user dimension is empty information, determining the supplementary information of the target user corresponding to the preset user dimension according to a preset feature filling mode;
and obtaining the feature information of each dimension contained by the target user according to the feature information and/or the supplementary information of the target user on each preset user dimension.
In the embodiment of the present disclosure, the feature information of each dimension included by the target user is obtained according to the feature information and/or the supplementary information of the target user in each preset user dimension, where each dimension feature information is obtained by determining corresponding feature information according to the configuration tag in each preset user dimension, or each dimension feature information is obtained by determining corresponding feature information according to the configuration tag in a part of each preset user dimension and obtaining each dimension feature information according to corresponding supplementary information in another part of each preset user dimension, or each dimension feature information is obtained according to corresponding supplementary information in each preset user dimension. Therefore, after the feature information of each preset user dimension is supplemented by the configuration label, all the preset user dimensions required by the launching condition can be matched during matching, so that a correct launching object corresponding to the configuration label is obtained, and the problem that the number of the matched launching objects is enlarged due to the lack of the feature information of each preset user dimension is solved.
Optionally, when a target casting crowd meeting the dimension characteristic information is not found from the casting crowd, the apparatus 40 further includes:
a third determining module, configured to determine, according to a seed user corresponding to the release object, an extended user similar to the seed user by using a preset similarity calculation method; the seed user is a user matched with the user portrait associated with the delivery object;
the selection module is used for selecting a plurality of users from the extended users and the seed users to form an extended release crowd according with the release plan number according to the release plan number configured by the release object;
and a fourth determining module, configured to determine, according to the dimensional feature information included in the target user, a target delivery crowd meeting the dimensional feature information from the extended delivery crowd.
In the embodiment of the disclosure, the preset similar calculation method may be a similar population expansion (lokalike) technology, and the similarity between other users and the seed user is calculated by the preset similar calculation method, so as to perform automatic similar population dynamic expansion, so that, on one hand, different requirements of different delivery plans on population orientation accuracy and coverage Rate can be met, and on the other hand, the Click Through Rate (Click-Through-Rate, CTR) and Conversion Rate (Conversion Rate, CVR) of the expanded population of the preset similar calculation method can be significantly improved, the distribution of intra-site traffic can be optimized, and the overflow price and the commercial change capability of the traffic can be improved.
In the embodiment of the present disclosure, according to each dimension feature information included in the target user, if the feature information on the dimension of the preset user appears as null information in the target user, the information on the dimension is supplemented in a preset feature filling manner, and then, a target release crowd meeting each dimension feature information is determined from the extended release crowd. Therefore, after the characteristic information on each dimension is supplemented, all the directional dimensions required by the task can be screened during screening, and a correct recall result is obtained.
Optionally, the object recommendation model includes a first recommendation model, a second recommendation model, and a third recommendation model, and the calculating module 402 is further configured to:
based on the configuration tag corresponding to the target user, the preset delivery condition on the target page and the attribute characteristics contained in the hit objects, obtaining user side characteristics corresponding to the target user, content side characteristics corresponding to the hit objects respectively and content related characteristics corresponding to each hit object by the target user;
for each hit object, inputting the user side features, the content-related features and the content side features into a pre-trained first recommendation model, and obtaining a first output value representing the preference degree of the target user for the hit object; inputting the user side characteristics to a pre-trained second recommendation model to obtain a second output value representing the user characteristics corresponding to the target user; inputting the content side features into a pre-trained third recommendation model to obtain a third output value representing the hot value of the hit object;
for each hit object, taking a difference value between a first output value and a bias output value corresponding to the hit object as a prediction score corresponding to the hit object; the offset output value is determined from a product of the second output value and the third output value;
and selecting the hit objects with the prediction scores meeting a preset score threshold from the hit objects as the objects to be released matched with the target user.
Optionally, the calculating module 402 is further configured to:
performing vector conversion and splicing on the user side features, the content related features and the content side features to obtain user feature vectors corresponding to the target user;
and inputting the user feature vector into the pre-trained first recommendation model to obtain an output result as the first output value.
Optionally, the apparatus 40 further includes:
the acquisition module is used for acquiring sample user characteristics, sample content related characteristics, sample content side characteristics and real scores of sample objects;
a fifth determining module, configured to use the sample user characteristic, the sample content-related characteristic, and the sample content-side characteristic as a first training sample, use the sample user characteristic as a second training sample, and use the sample content-side characteristic as a third training sample;
the training module is used for performing iterative training on an initial first recommendation model by using the first training sample, performing iterative training on an initial second recommendation model by using the second training sample, and performing iterative training on an initial third recommendation model by using the third training sample to obtain the pre-trained first recommendation model, the pre-trained second recommendation model and the pre-trained third recommendation model of which the output training scores accord with the real scores.
In the embodiment of the disclosure, a first training sample may be input to an initial first recommendation model for calculation, a second training sample may be input to an initial second recommendation model for calculation, a third training sample may be input to an initial third recommendation model for calculation, a training prediction score may be calculated by a preset prediction score formula, a loss function between the training prediction score and a true score may be calculated, if a result of the loss function does not meet a preset threshold, training parameters in the first recommendation model, the second recommendation model, and the third recommendation model may be adjusted, and an operation of inputting the training sample to the initial recommendation model for calculation may be re-executed until a result of the loss function meets the preset threshold; if the loss function result meets the preset threshold value, that is, it is determined that the training prediction score is approximate to the real score, the first recommendation model obtained through training may be used as a pre-training first recommendation model, the second recommendation model obtained through training may be used as a pre-training second recommendation model, and the second recommendation model obtained through training may be used as a pre-training second recommendation model. Therefore, the model can learn the capability of correctly calculating and processing the sample characteristics to obtain the real prediction score by continuously adopting the sample characteristics for training and adjusting the training parameters of the model.
Optionally, the second determining module 403 is further configured to:
for the plurality of objects to be put, determining the exposure time and the exposure times corresponding to each object to be put;
and selecting the object to be launched of which the exposure time and the exposure times are smaller than a preset exposure threshold value as the target launched object according to the exposure time and the exposure times corresponding to each object to be launched.
Optionally, after the target delivery object is pushed to the target user on the target page in the preset delivery manner, the apparatus 40 further includes:
the monitoring module is used for monitoring feedback information of the target user to the target delivery object; and if the feedback information does not accord with a preset feedback threshold value, adjusting the preset delivery strategy, and reselecting the target delivery object from the plurality of objects to be delivered according to the adjusted delivery strategy.
In the embodiment of the disclosure, in order to ensure the effect of the delivery system, monitoring feedback is particularly important, on one hand, an effective evaluation can be performed on the content quality of the delivered object, and on the other hand, the influence of the delivery plan on the flow distribution efficiency can also be effectively observed, so that feedback information of a target user on the target delivered object can be monitored, if the feedback information does not meet a preset feedback threshold, a preset delivery strategy is adjusted, and the target delivered object is reselected from a plurality of objects to be delivered according to the adjusted delivery strategy, where the feedback information may be service indexes such as a click rate and a conversion rate of the user on the delivered object.
Having described the object delivery method and apparatus according to the exemplary embodiment of the present disclosure, a storage medium according to the exemplary embodiment of the present disclosure will be described with reference to fig. 9.
Referring to fig. 9, a storage medium 500 for implementing the above method according to an embodiment of the present disclosure is described, which may employ a portable compact disc read only memory (CD-ROM) and include program code, and may be run on a device, such as a personal computer. However, the program product of the present disclosure is not so limited, and in this document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
A computer readable signal medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Program code for carrying out operations of the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user computing device, partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., through the internet using an internet service provider).
Having described the storage medium of the exemplary embodiment of the present disclosure, next, an electronic device of the exemplary embodiment of the present disclosure will be explained with reference to fig. 10.
The electronic device 600 shown in fig. 10 is only an example and should not bring any limitations to the function and scope of use of the embodiments of the present disclosure.
As shown in fig. 10, the electronic device 600 is embodied in the form of a general purpose computing device. The components of the electronic device 600 may include, but are not limited to: the at least one processing unit 610, the at least one memory unit 620, a bus 630 connecting different system components (including the memory unit 620 and the processing unit 610), and a display unit 640.
Wherein the storage unit stores program code that is executable by the processing unit 610 to cause the processing unit 610 to perform steps according to various exemplary embodiments of the present disclosure as described in the above section "exemplary methods" of this specification. For example, the processing unit 610 may execute step S101, in response to an object placement request of a target user on a target page, and according to a configuration tag corresponding to the target user and a placement condition configured in advance on the target page, determine a plurality of hit objects corresponding to the placement condition that meets the configuration tag; step S102, calculating the corresponding prediction scores of the hit objects by using a pre-trained object recommendation model to determine a plurality of objects to be delivered matched with the target user; the object recommendation model is a model for determining the target user's preference for the hit object, the target user's characteristics, and the hit object's heat; step S103, determining a target release object corresponding to the target user from the plurality of objects to be released according to a preset release strategy; and step S104, pushing the target delivery object to the target user on the target page in a preset delivery mode.
The storage unit 620 may include volatile storage, such as Random Access Memory (RAM) 6201 and/or cache memory unit 6202, and may further include read-only memory (ROM) 6203.
The memory unit 620 may also include a program/utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.
The bus 630 may include a data bus, an address bus, and a control bus.
The electronic device 600 may also communicate with one or more external devices 70 (e.g., keyboard, pointing device, bluetooth device, etc.) via an input/output (I/O) interface 650. The electronic device 600 further comprises a display unit 640 connected to the input/output (I/O) interface 650 for displaying. Also, the electronic device 600 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network such as the Internet) via the network adapter 660. As shown, the network adapter 660 communicates with the other modules of the electronic device 600 over the bus 630. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, to name a few.
It should be noted that although in the above detailed description several modules or sub-modules of the object delivery device and the audio sharing device are mentioned, such a division is merely exemplary and not mandatory. Indeed, the features and functionality of two or more of the units/modules described above may be embodied in one unit/module, in accordance with embodiments of the present disclosure. Conversely, the features and functions of one unit/module described above may be further divided into embodiments by a plurality of units/modules.
Through the description of the foregoing embodiments, it is clear to those skilled in the art that the method of the foregoing embodiments may be implemented by software plus a necessary general hardware platform, and certainly may also be implemented by hardware, but in many cases, the former is a better implementation. Based on such understanding, the technical solutions of the present disclosure may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal (e.g., a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present disclosure.
While the embodiments of the present disclosure have been described in connection with the appended drawings, the present disclosure is not limited to the specific embodiments, which have been described above for illustrative purposes only and not for purposes of limitation, and it will be appreciated by those of ordinary skill in the art that, in light of the present disclosure, numerous modifications may be made without departing from the spirit of the disclosure and scope of the appended claims.

Claims (10)

1. An object delivery method, the method comprising:
responding to an object launching request of a target user on a target page, and determining a plurality of hit objects corresponding to launching conditions of configuration tags according to the configuration tags corresponding to the target user and launching conditions pre-configured on the target page;
calculating the prediction scores corresponding to the hit objects by using a pre-trained object recommendation model to determine a plurality of objects to be delivered matched with the target user; the object recommendation model is a model for determining the target user's preference for the hit object, the target user's characteristics, and the hit object's heat;
determining a target delivery object corresponding to the target user from the plurality of objects to be delivered according to a preset delivery strategy;
and pushing the target delivery object to the target user on the target page in a preset delivery mode.
2. The method according to claim 1, wherein the pre-configured serving conditions include a serving crowd and a serving object corresponding to the crowd, and the determining, according to the configuration tag corresponding to the target user and the serving conditions pre-configured on the target page, a plurality of hit objects corresponding to the serving conditions that meet the configuration tag includes:
determining all dimension characteristic information contained by the target user based on the configuration label corresponding to the target user;
determining a target release crowd meeting the dimension characteristic information from the release crowd according to the dimension characteristic information contained in the target user;
and determining a plurality of throwing objects corresponding to the target throwing crowd as a plurality of hitting objects corresponding to the throwing conditions meeting the configuration label based on a preset mapping relation between the throwing crowd and the throwing objects.
3. The method according to claim 2, wherein the determining, based on the configuration tag corresponding to the target user, the dimensional feature information included in the target user includes:
determining feature information corresponding to each preset user dimension of the target user according to the configuration label corresponding to the target user;
if the feature information of the configuration label corresponding to the preset user dimension is empty information, determining the supplementary information of the target user corresponding to the preset user dimension according to a preset feature filling mode;
and obtaining the feature information of each dimension contained by the target user according to the feature information and/or the supplementary information of the target user on each preset user dimension.
4. The method of claim 2, wherein when a target cast population matching the dimensional characteristics is not found from the cast population, the method further comprises:
determining an extended user similar to the seed user by using a preset similarity calculation method according to the seed user corresponding to the release object; the seed user is a user matched with the user portrait associated with the delivery object;
selecting a plurality of users from the extended users and the seed users to form an extended release crowd according with the release plan number according to the release plan number configured by the release object;
and determining a target release crowd meeting the dimension characteristic information from the extended release crowd according to the dimension characteristic information contained in the target user.
5. The method according to claim 1, wherein the object recommendation model comprises a first recommendation model, a second recommendation model and a third recommendation model, and the determining the target user matching multiple objects to be delivered by calculating the prediction scores corresponding to the multiple hit objects by using a pre-trained object recommendation model comprises:
based on the configuration tag corresponding to the target user, the preset delivery condition on the target page and the attribute characteristics contained in the hit objects, obtaining user side characteristics corresponding to the target user, content side characteristics corresponding to the hit objects respectively and content related characteristics corresponding to each hit object by the target user;
for each hit object, inputting the user side features, the content-related features and the content side features into a pre-trained first recommendation model, and obtaining a first output value representing the preference degree of the target user for the hit object; inputting the user side characteristics to a pre-trained second recommendation model to obtain a second output value representing the user characteristics corresponding to the target user; inputting the content side features into a pre-trained third recommendation model to obtain a third output value representing the hot value of the hit object;
for each hit object, taking the difference value between the first output value and the offset output value corresponding to the hit object as the prediction score corresponding to the hit object; the offset output value is determined from a product of the second output value and the third output value;
and selecting the hit object with the prediction score meeting a preset score threshold value from the hit objects as the object to be launched matched with the target user.
6. The method according to claim 1, wherein the determining a target delivery object corresponding to the target user from the plurality of objects to be delivered according to a preset delivery policy comprises:
determining the exposure time and the exposure times corresponding to each object to be launched for the plurality of objects to be launched;
and selecting the object to be launched of which the exposure time and the exposure times are smaller than a preset exposure threshold value as the target launched object according to the exposure time and the exposure times corresponding to each object to be launched.
7. The method according to claim 1, wherein after the target delivery object is pushed to the target user on the target page in a preset delivery manner, the method further comprises:
monitoring feedback information of the target user to the target delivery object;
and if the feedback information does not accord with a preset feedback threshold value, adjusting the preset delivery strategy, and reselecting the target delivery object from the plurality of objects to be delivered according to the adjusted delivery strategy.
8. An object delivery apparatus, the apparatus comprising:
the system comprises a first determining module, a second determining module and a third determining module, wherein the first determining module is used for responding to an object launching request of a target user on a target page, and determining a plurality of hit objects corresponding to launching conditions of configuration tags according to the configuration tags corresponding to the target user and launching conditions configured in advance on the target page;
the calculation module is used for calculating the prediction scores corresponding to the hit objects by using a pre-trained object recommendation model to determine a plurality of objects to be delivered matched with the target user; the object recommendation model is a model for determining the target user's preference for the hit object, the target user's characteristics, and the hit object's heat;
the second determining module is used for determining a target release object corresponding to the target user from the plurality of objects to be released according to a preset release strategy;
and the pushing module is used for pushing the target delivery object to the target user on the target page in a preset delivery mode.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the object delivery method according to any one of claims 1 to 7.
10. An electronic device, comprising:
a processor; and
a memory for storing executable instructions of the processor;
wherein the processor is configured to perform the object delivery method of any of claims 1-7 via execution of the executable instructions.
CN202211314115.9A 2022-10-25 2022-10-25 Object delivery method and device, storage medium and electronic equipment Pending CN115601088A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211314115.9A CN115601088A (en) 2022-10-25 2022-10-25 Object delivery method and device, storage medium and electronic equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211314115.9A CN115601088A (en) 2022-10-25 2022-10-25 Object delivery method and device, storage medium and electronic equipment

Publications (1)

Publication Number Publication Date
CN115601088A true CN115601088A (en) 2023-01-13

Family

ID=84849906

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211314115.9A Pending CN115601088A (en) 2022-10-25 2022-10-25 Object delivery method and device, storage medium and electronic equipment

Country Status (1)

Country Link
CN (1) CN115601088A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115934809A (en) * 2023-03-08 2023-04-07 北京嘀嘀无限科技发展有限公司 Data processing method and device and electronic equipment

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115934809A (en) * 2023-03-08 2023-04-07 北京嘀嘀无限科技发展有限公司 Data processing method and device and electronic equipment

Similar Documents

Publication Publication Date Title
CN109104620B (en) Short video recommendation method and device and readable medium
CN101601026B (en) System and method for effectively providing content to client devices in an electronic network
CN108960316B (en) Method and apparatus for generating a model
JP2010250827A (en) Touchpoint customization system
CN105493057A (en) Content selection with precision controls
CN105453122A (en) Contextual mobile application advertisements
CN108491540B (en) Text information pushing method and device and intelligent terminal
CN110674349B (en) Video POI (Point of interest) identification method and device and electronic equipment
CN113742567B (en) Recommendation method and device for multimedia resources, electronic equipment and storage medium
CN104486649A (en) Video content rating method and device
CN114860892B (en) Hierarchical category prediction method, device, equipment and medium
CN113393299A (en) Recommendation model training method and device, electronic equipment and storage medium
CN115601088A (en) Object delivery method and device, storage medium and electronic equipment
WO2016099447A1 (en) Personalized application recommendations
CN108959450B (en) Popular comment determination system, method, device and readable medium
US20240046922A1 (en) Systems and methods for dynamically updating machine learning models that provide conversational responses
GB2600817A (en) Systems and methods for generating dynamic interface options using machine learning models
US20150170035A1 (en) Real time personalization and categorization of entities
CN114580790A (en) Life cycle stage prediction and model training method, device, medium and equipment
CN115017362A (en) Data processing method, electronic device and storage medium
CN115344774A (en) User account screening method and device and server
CN111339291B (en) Information display method and device and storage medium
CN111797273A (en) Method and device for adjusting parameters
CN112288452A (en) Advertisement preview method and device, electronic equipment and storage medium
CN114595323B (en) Portrait construction, recommendation, model training method, apparatus, device and storage medium

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