CN116796053A

CN116796053A - Resource pushing method, device, computer equipment and medium

Info

Publication number: CN116796053A
Application number: CN202210245342.4A
Authority: CN
Inventors: 袁坤
Original assignee: Beijing Dajia Internet Information Technology Co Ltd
Current assignee: Beijing Dajia Internet Information Technology Co Ltd
Priority date: 2022-03-14
Filing date: 2022-03-14
Publication date: 2023-09-22

Abstract

The disclosure relates to a resource pushing method, a resource pushing device, computer equipment and a resource pushing medium, and belongs to the technical field of Internet. In the embodiment of the disclosure, when pushing resources for a target object, the weight coefficient corresponding to each behavior dimension is determined based on the first parameter and the second parameter of the target object in each behavior dimension, so when determining the weight coefficient corresponding to each behavior dimension, not only the difference of the target object relative to the object population in each behavior dimension but also the distribution condition of a plurality of candidate resources of the target object in each behavior dimension are referred to, and the efficiency of determining the weight coefficient is improved, and meanwhile, the accuracy of determining the weight coefficient is improved, and further, the pushing parameters of the candidate resources are determined based on the behavior prediction parameters of the candidate resources of the target object in each behavior dimension and the weight coefficients corresponding to each behavior dimension, and then the resource pushing is performed based on the pushing parameters, so that the efficiency and the accuracy of the resource pushing are improved.

Description

Resource pushing method, device, computer equipment and medium

Technical Field

The disclosure relates to the technical field of internet, and in particular relates to a resource pushing method, a resource pushing device, computer equipment and a medium.

Background

With the rapid development of internet technology and the gradual expansion of the scale of network users, internet-type multimedia resources have penetrated into aspects of people's life, and simultaneously, the pushing method of multimedia resources has also rapidly developed. In general, when pushing a multimedia resource to a network user, multiple types of interaction actions that occur for other multimedia resources, such as a positive interaction action like a clicking action, a collecting action, and the like, a negative interaction action like a shielding action, a skipping action, and the like, and a viewing duration for other multimedia resources, all affect a pushing result of the multimedia resource. Therefore, the pushing problem of the multimedia resource is a multi-objective (i.e. multi-behavior dimension) optimization problem, and the optimization purpose is to promote the forward interaction behavior of the user as much as possible on the basis of ensuring the viewing time of the multimedia resource.

At present, when the above-mentioned multi-objective optimization problem is faced, it is required to balance the relationship between each behavior dimension, usually, a service personnel sets a weight coefficient for each behavior dimension according to historical experience, and then determines a pushing parameter of a multimedia resource based on a pushing sub-parameter of each behavior dimension and a weight coefficient corresponding to each behavior dimension, and then pushes the multimedia resource to a network user based on the pushing parameter of the multimedia resource.

However, relying on human determination in the above process reduces the efficiency and accuracy of resource pushing.

Disclosure of Invention

The disclosure provides a resource pushing method, a device, computer equipment and a medium, which improve the efficiency of determining a weight coefficient and the accuracy of determining the weight coefficient, thereby improving the efficiency and accuracy of resource pushing. The technical scheme of the present disclosure is as follows:

according to a first aspect of an embodiment of the present disclosure, there is provided a resource pushing method, including:

acquiring behavior prediction parameters of candidate resources of a target object in at least two behavior dimensions, wherein the behavior prediction parameters represent the prediction condition of interactive behaviors of the target object and the candidate resources in the behavior dimensions;

determining a weight coefficient corresponding to the behavior dimension based on a first parameter and a second parameter of the target object in the behavior dimension, wherein the first parameter represents the difference of the target object relative to an object group in the behavior dimension, and the second parameter represents the distribution condition of a plurality of candidate resources of the target object in the behavior dimension;

and determining a pushing parameter of the candidate resource based on the behavior prediction parameters of the candidate resource in the at least two behavior dimensions and the weight coefficients corresponding to the at least two behavior dimensions, wherein the pushing parameter is used for determining the pushing sequence of the candidate resource when the candidate resource is pushed for the target object.

In some embodiments, the acquiring the first parameter includes:

acquiring a first average value of the target object in the behavior dimension and a second average value of the object group in the behavior dimension, wherein the first average value is an average value of behavior prediction parameters of candidate resources of the target object in the behavior dimension, and the second average value is an average value of behavior prediction parameters of candidate resources of a plurality of objects included in the object group in the behavior dimension;

the first parameter is determined based on the first average value and the second average value, wherein the first parameter is positively correlated with the first average value and the first parameter is negatively correlated with the second average value.

In some embodiments, the acquiring of the second parameter includes:

and acquiring dispersion of behavior prediction parameters of the plurality of candidate resources of the target object in the behavior dimension, and determining the dispersion as the second parameter.

In some embodiments, determining the weight coefficient corresponding to the behavior dimension based on the first parameter and the second parameter of the target object in the behavior dimension comprises:

determining a third parameter of the target object in the behavior dimension based on the first parameter and the second parameter of the target object in the behavior dimension, wherein the third parameter is positively correlated with the first parameter and the third parameter is positively correlated with the second parameter;

And determining a weight coefficient corresponding to the behavior dimension based on the initial weight coefficient corresponding to the behavior dimension and the third parameter, wherein the weight coefficient is positively correlated with the initial weight coefficient, and the weight coefficient is positively correlated with the third parameter.

In some embodiments, determining a third parameter of the target object in the behavioral dimension based on the first parameter and the second parameter of the target object in the behavioral dimension comprises:

determining a fourth parameter of the target object in the behavior dimension based on a second parameter of the target object in the behavior dimension, the fourth parameter being inversely related to the second parameter;

the third parameter is determined based on the first parameter and the fourth parameter of the target object in the behavior dimension, the third parameter being inversely related to the fourth parameter.

In some embodiments, obtaining behavior prediction parameters for candidate resources of the target object in at least two behavior dimensions comprises:

and determining a behavior prediction parameter of the candidate resource in the at least two behavior dimensions based on the object characteristic information of the target object and the content characteristic information of the candidate resource.

In some embodiments, determining the behavior prediction parameters of the candidate resource in the at least two behavior dimensions based on the object characteristic information of the target object and the content characteristic information of the candidate resource comprises:

And inputting the object characteristic information of the target object and the content characteristic information of the candidate resource into a behavior prediction model corresponding to the behavior dimension, predicting the behavior prediction parameter of the candidate resource in the behavior dimension through the behavior prediction model to obtain the behavior prediction parameter of the candidate resource in the behavior dimension, wherein the behavior prediction model is obtained by training based on the object characteristic information of the sample object, the content characteristic information of the candidate resource of the sample object and the behavior parameter label of the candidate resource of the sample object in the behavior dimension.

In some embodiments, determining the push parameters for the candidate resource based on the weight coefficients for the behavior prediction parameters for the candidate resource in the at least two behavior dimensions corresponding to the at least two behavior dimensions comprises:

sequencing the candidate resources in the at least two behavior dimensions according to behavior prediction parameters of the candidate resources in the at least two behavior dimensions;

determining a pushing subparameter of the candidate resource in the at least two behavior dimensions based on the arrangement order of the candidate resource in the at least two behavior dimensions, wherein the pushing subparameter is inversely related to the arrangement order;

And determining the pushing parameters of the candidate resource based on the pushing sub-parameters of the candidate resource in the at least two behavior dimensions and the weight coefficients corresponding to the at least two behavior dimensions.

In some embodiments, in the at least two behavior dimensions, ordering according to the behavior prediction parameters of the candidate resource in the at least two behavior dimensions, respectively, includes:

and sequencing according to the behavior prediction parameters of the candidate resource in the at least two behavior dimensions in the queues corresponding to the at least two behavior dimensions respectively.

In some embodiments, determining the push parameters of the candidate resource based on the push sub-parameters of the candidate resource in the at least two behavior dimensions and the weight coefficients corresponding to the at least two behavior dimensions comprises:

and carrying out weighted summation based on the pushing sub-parameters of the candidate resource in the at least two behavior dimensions and the weight coefficients corresponding to the at least two behavior dimensions to obtain the pushing parameters of the candidate resource.

According to a second aspect of the embodiments of the present disclosure, there is provided a resource pushing device, including:

an obtaining unit configured to perform obtaining a behavior prediction parameter of a candidate resource of a target object in at least two behavior dimensions, the behavior prediction parameter representing a predicted situation in which the target object interacts with the candidate resource in the behavior dimension;

A weight coefficient determination unit configured to perform determination of a weight coefficient corresponding to the behavior dimension based on a first parameter and a second parameter of the target object in the behavior dimension, the first parameter representing a difference of the target object in the behavior dimension relative to an object population, the second parameter representing a distribution of a plurality of the candidate resources of the target object in the behavior dimension;

and a pushing parameter determining unit configured to determine a pushing parameter of the candidate resource based on the weight coefficients of the behavior prediction parameter of the candidate resource in the at least two behavior dimensions and the at least two behavior dimensions, wherein the pushing parameter is used for determining a pushing order of the candidate resource when the candidate resource is pushed for the target object.

In some embodiments, the weight coefficient determination unit includes:

a first obtaining subunit configured to perform obtaining a first average value of the target object in the behavior dimension, where the first average value is an average value of behavior prediction parameters of candidate resources of the target object in the behavior dimension, and a second average value of behavior prediction parameters of candidate resources of a plurality of objects included in the object group in the behavior dimension;

A determining subunit configured to perform determining the first parameter based on the first average and the second average, wherein the first parameter is positively correlated with the first average and the first parameter is negatively correlated with the second average.

In some embodiments, the weight coefficient determination unit includes:

a second acquisition subunit configured to perform acquiring a dispersion of the behavior prediction parameters of the plurality of candidate resources of the target object in the behavior dimension, determining the dispersion as the second parameter.

In some embodiments, the weight coefficient determination unit includes:

a third parameter determination subunit configured to perform determining a third parameter of the target object in the behavior dimension based on the first parameter and the second parameter of the target object in the behavior dimension, wherein the third parameter is positively correlated with the first parameter and the third parameter is positively correlated with the second parameter;

and a weight coefficient determination subunit configured to determine a weight coefficient corresponding to the behavior dimension based on the initial weight coefficient corresponding to the behavior dimension and the third parameter, wherein the weight coefficient is positively correlated with the initial weight coefficient, and the weight coefficient is positively correlated with the third parameter.

In some embodiments, the third parameter determination subunit is configured to perform:

In some embodiments, the acquisition unit is configured to perform:

In some embodiments, the push parameter determination unit comprises:

a ranking subunit configured to perform ranking in the at least two behavior dimensions according to the behavior prediction parameters of the candidate resource in the at least two behavior dimensions, respectively;

a determining subunit configured to perform determining, based on an order of arrangement of the candidate resource in the at least two behavior dimensions, push sub-parameters of the candidate resource in the at least two behavior dimensions, respectively, the push sub-parameters being inversely related to the order of arrangement;

the determining subunit is further configured to determine a pushing parameter of the candidate resource based on the pushing sub-parameter of the candidate resource in the at least two behavior dimensions and the weight coefficients corresponding to the at least two behavior dimensions.

In some embodiments, the ordering subunit is configured to perform:

In some embodiments, the determining subunit is further configured to perform:

According to a third aspect of embodiments of the present disclosure, there is provided a computer device comprising:

one or more processors;

a memory for storing the processor-executable program code;

wherein the processor is configured to execute the program code to implement the resource pushing method described above.

According to a fourth aspect of embodiments of the present disclosure, there is provided a computer-readable storage medium comprising: the program code in the computer readable storage medium, when executed by a processor of a computer device, enables the computer device to perform the resource pushing method described above.

According to a fifth aspect of embodiments of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the resource pushing method described above.

According to the technical scheme provided by the embodiment of the disclosure, when the resource is pushed for the target object, the weight coefficient corresponding to each behavior dimension is determined based on the first parameter and the second parameter of the target object in each behavior dimension, so that when the weight coefficient corresponding to each behavior dimension is determined, not only is the difference of the target object relative to the object group in each behavior dimension referred to, but also the distribution condition of a plurality of candidate resources of the target object in each behavior dimension referred to, the determined weight coefficient can embody the tendency degree of the target object in each behavior dimension and the distribution characteristic of the candidate resources in each behavior dimension, and the weight coefficient corresponding to each behavior dimension can be further embodied, and according to the first parameter and the second parameter of each object in different behavior dimensions, the weight coefficient of each object in different behavior dimensions is determined, so that the different weight coefficient can be determined according to the different objects, the characteristic of each object and the weighted weight coefficient of each resource can be embodied, the weight coefficient of each candidate object is improved, the weight coefficient of each candidate object is further improved, and the pushing resource is further predicted based on the weight coefficient of each behavior dimension, and the accuracy is further improved, and the pushing resource is predicted based on the candidate performance parameters.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure and do not constitute an undue limitation on the disclosure.

FIG. 1 is a schematic diagram of an implementation environment of a resource pushing method according to an exemplary embodiment;

FIG. 2 is a flowchart illustrating a method of pushing resources according to an example embodiment;

FIG. 3 is a flowchart illustrating a method of pushing resources according to an example embodiment;

FIG. 4 is a flowchart illustrating a method of pushing resources according to an example embodiment;

FIG. 5 is a block diagram of a resource pushing device, according to an example embodiment;

FIG. 6 is a block diagram of a terminal shown in accordance with an exemplary embodiment;

fig. 7 is a block diagram of a server, according to an example embodiment.

Detailed Description

In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the foregoing figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the disclosure described herein may be capable of operation in sequences other than those illustrated or described herein. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.

The data or information to which the present disclosure relates may be data or information that is authorized by a user or sufficiently authorized by parties.

First, description is made for an application scenario related to an embodiment of the present disclosure:

the resource pushing method provided by the embodiment of the disclosure can be applied to pushing scenes of multimedia resources, such as short video pushing scenes. In some embodiments, when facing the multi-objective (i.e., multi-behavior dimension) optimization problem in the push scenario, the resource push method provided by the embodiments of the present disclosure may be used to balance the relationship between each behavior dimension, such as balancing the relationship between the viewing duration of the multimedia resource and the forward interaction behavior.

It should be noted that, the pushing process of the multimedia resource generally needs to go through three stages of recall, coarse arrangement and fine arrangement. In some embodiments, the resource pushing method provided by the embodiments of the present disclosure is applied in a fine-ranking stage of resource pushing.

The recall stage is a first stage of pushing, specifically, according to the user characteristics and the resource characteristics, a part of multimedia resources potentially interested by the user are quickly retrieved from a massive multimedia resource library, and the part of multimedia resources are added into a recall queue. The coarse ranking stage is a second pushing stage, specifically, the multimedia resources in the recall queue are ranked according to some coarse ranking indexes (such as click rate) so as to screen out multimedia resources with the front ranking, and the screened multimedia resources are added into the coarse ranking queue so as to reduce the number of recalled resources, thereby reducing the ranking pressure of the fine ranking stage. The fine ranking stage is a third stage of pushing, specifically, according to ranking indexes (such as thousands of times of revenues) of fine ranking, the multimedia resources in the coarse ranking queue are further ranked to obtain a fine ranking queue, so that the multimedia resources with the forefront ranking are screened out for pushing.

Fig. 1 is a schematic diagram of an implementation environment of a resource pushing method provided by an embodiment of the present disclosure, referring to fig. 1, the implementation environment includes: the server 101.

The server 101 may be an independent physical server, a server cluster or a distributed file system formed by a plurality of physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs (Content Delivery Network, content delivery networks), basic cloud computing services such as big data and artificial intelligence platforms, and the like.

In some embodiments, the server 101 is provided with functionality to push multimedia assets for users. In this embodiment of the disclosure, the server 101 is configured to obtain a behavior prediction parameter of a candidate resource of a target object in at least two behavior dimensions, determine a weight coefficient corresponding to the behavior dimension based on a first parameter and a second parameter of the target object in the behavior dimension, and determine a pushing parameter of the candidate resource based on a weight coefficient corresponding to the behavior prediction parameter of the candidate resource in the at least two behavior dimensions and the at least two behavior dimensions. The disclosed embodiments subsequently employ the target object to refer to the user of the multimedia asset to be pushed.

In some embodiments, the number of servers 101 may be greater or lesser, which is not limited by the disclosed embodiments. Of course, the server 101 may also include other functional servers in order to provide more comprehensive and diverse services.

In some embodiments, the implementation environment further comprises: a terminal 102. In some embodiments, the server 101 is directly or indirectly connected to the terminal 102 through wired or wireless communication, which is not limited by the embodiments of the present disclosure.

The terminal 102 may be at least one of a smart phone, a smart watch, a desktop computer, a laptop computer, a virtual reality terminal, an augmented reality terminal, a wireless terminal, and a laptop portable computer. The terminal 102 has a communication function and can access a wired network or a wireless network. The terminal 102 may refer broadly to one of a plurality of terminals, the present embodiment being illustrated by the terminal 102 only. Those skilled in the art will recognize that the number of terminals may be greater or lesser.

In some embodiments, the terminal 102 is running an application with multimedia asset playing capabilities, such as a social application, a live application, a short video application, and so on. In some embodiments, server 101 provides background services for applications run by terminal 102. In the embodiment of the present disclosure, the terminal 102 is configured to send a resource pushing request to the server 101 in response to the resource pushing request triggered by the target object, so as to trigger the server 101 to push the multimedia resource for the target object.

FIG. 2 is a flowchart illustrating a method of pushing resources, as shown in FIG. 2, performed by a server, according to an exemplary embodiment, comprising the steps of:

in step 201, the server obtains behavior prediction parameters of a candidate resource of a target object in at least two behavior dimensions, where the behavior prediction parameters represent a predicted situation of interaction behavior of the target object with the candidate resource in the behavior dimensions.

In step 202, the server determines a weight coefficient corresponding to the behavior dimension based on a first parameter and a second parameter of the target object in the behavior dimension, where the first parameter represents a difference of the target object in the behavior dimension relative to a population of objects, and the second parameter represents a distribution of the plurality of candidate resources of the target object in the behavior dimension.

In step 203, the server determines a pushing parameter of the candidate resource based on the weight coefficients of the behavior prediction parameter of the candidate resource in the at least two behavior dimensions and the at least two behavior dimensions, where the pushing parameter is used to determine a pushing order of the candidate resource when pushing the resource for the target object.

According to the technical scheme, when the resources are pushed for the target object, the weight coefficient corresponding to each behavior dimension is determined based on the first parameter and the second parameter of the target object in each behavior dimension, so that when the weight coefficient corresponding to each behavior dimension is determined, not only is the difference of the target object relative to the object group in each behavior dimension referred to, but also the distribution condition of a plurality of candidate resources of the target object in each behavior dimension referred to, the determined weight coefficient can embody the tendency degree of the target object in each behavior dimension and the distribution characteristic of the candidate resources in each behavior dimension referred to, and according to the first parameter and the second parameter of each object in different behavior dimensions, the weight coefficient of each object in different behavior dimensions is determined, so that different weight coefficients can be determined according to different objects, the personalized weight coefficient of the object characteristics and the resource characteristics of each object can be determined, the weight coefficient is determined, the weight coefficient of the target object is improved, the distribution characteristics of the candidate resources in each behavior dimension can be improved, and the pushing resources are predicted based on the weight coefficient of each behavior dimension, and the pushing resources are predicted based on the corresponding behavior parameters.

In some embodiments, the acquiring the first parameter includes:

In some embodiments, the acquiring of the second parameter includes:

The foregoing fig. 2 is merely a basic flow of the disclosure, and the following further describes a scheme provided by the disclosure based on a specific embodiment, and fig. 3 is a flowchart of a resource pushing method according to an exemplary embodiment, and referring to fig. 3, the method includes:

In step 301, the terminal sends a resource push request of a target object to the server, where the resource push request is used to request the server to push a resource for the target object.

The terminal is a terminal corresponding to the target object. The target object refers to a target user, i.e. a user who pushes a multimedia resource, i.e. a user who is to push the multimedia resource. In some embodiments, the target object is represented by an object identification, which may be, for example, an object name, an object account number, an object ID (Identity), or the like.

In the embodiment of the disclosure, the terminal runs an application program with a multimedia resource playing function, such as a social application program, a live broadcast application program, a short video application program and the like. Correspondingly, the server is a background server of the application program with the multimedia resource playing function. In some embodiments, the multimedia asset is any one of a video, a picture, or an article. In some embodiments, the multimedia assets are represented by asset identifications, which may be asset names, asset numbers, asset IDs, and the like, for example.

In some embodiments, the resource pushing request is triggered based on a start operation of the target object on the application program with the multimedia resource playing function, for example, the target object operates on the terminal, and starts the application program with the multimedia resource playing function, and then the terminal responds to the start operation of the target object on the application program, sends the resource pushing request of the target object to the server, so as to request the server to push the multimedia resource for the target object; or, the resource pushing request is triggered based on a resource switching operation of the target object in the application program, for example, if the target object performs the switching operation on the currently displayed multimedia resource in the process of viewing the multimedia resource in the application program, the terminal responds to the switching operation of the target object on the multimedia resource, and sends the resource pushing request to the server to request the server to push the multimedia resource for the target object. Of course, the resource push request can also be triggered based on other operations of the target object, such as an access operation of the target object to a push page in the application program or a refresh operation of the target object in the application program or a search operation of the target object in the application program, which is not limited by the embodiments of the present disclosure.

In step 302, the server responds to a resource pushing request of a target object, and obtains behavior prediction parameters of candidate resources of the target object in at least two behavior dimensions, where the behavior prediction parameters represent prediction situations of interaction behaviors of the target object with the candidate resources in the behavior dimensions.

The resource pushing request carries an object identifier of the target object. In some embodiments, after receiving the resource push request of the target object, the server obtains an object identifier of the target object carried by the resource push request, obtains a candidate resource of the target object based on the object identifier of the target object, and further obtains a behavior prediction parameter of the candidate resource in at least two behavior dimensions based on the candidate resource of the target object.

In the embodiment of the present disclosure, the candidate resource refers to a candidate multimedia resource. In some embodiments, the number of candidate resources is a plurality. In some embodiments, the candidate resources refer to a plurality of multimedia resources included in the coarse queuing queue, and further in the fine queuing stage of pushing, a subsequent resource pushing process is executed based on the plurality of multimedia resources included in the coarse queuing queue in combination with the resource pushing method provided by the embodiments of the present disclosure.

In the embodiment of the disclosure, the behavior dimension refers to dimensions corresponding to different types of interaction behaviors. In some embodiments, the behavior dimension includes at least two of a click dimension (click), a like dimension (like), a focus dimension (follow), a comment dimension (comment), an access personal page dimension (enter profile), a collection dimension, a mask dimension, a skip dimension, and a view duration dimension of the multimedia asset. Accordingly, the behavior prediction parameters include at least two of click rate, praise rate, attention rate, comment rate, access rate of individual pages, collection rate, mask rate, skip rate, and predicted viewing duration. Wherein, the click rate represents the probability of clicking the candidate resource by the target object; the praise rate represents the probability of praise of the target object for the candidate resource; the attention rate represents the probability that the target object is paying attention to the candidate resource; the comment rate represents the probability that the target object comments on the candidate resource; the access rate of the individual page represents the probability that the target object accesses the individual page associated with the candidate resource; the collection rate represents the probability that the target object is collected for the candidate resource; the mask rate represents the probability that the target object masks the candidate resource; the skip rate represents the probability that the target object skips for the candidate resource; the predicted viewing duration represents a predicted duration for which the target object is viewing the candidate asset. The above embodiments illustrate various types of behavior dimensions, however, in other embodiments, the behavior dimensions may include other types of behavior dimensions, such as a forwarding dimension, a number of views dimension of a multimedia resource, and the like, which are not limited by the embodiments of the present disclosure.

In some embodiments, the server determines the behavior prediction parameters of the candidate resource in the at least two behavior dimensions based on object characteristic information of the target object and content characteristic information of the candidate resource.

Wherein the object feature information represents object features of the target object. In some embodiments, the object feature information includes attribute feature information and behavior feature information of the target object, wherein the attribute feature information represents an attribute feature of the target object, such as gender, age, city, hobbies, and the like, and the behavior feature information represents a behavior feature of the target object, such as click, search, praise, collection, and the like. In some embodiments, the server obtains the attribute feature information and the behavior feature information based on a user representation of the target object, the user representation describing the attribute information and the behavior information of the object. The content characteristic information indicates content characteristics of the candidate resource. In some embodiments, the content characteristic information is a text characteristic or an image characteristic of the candidate resource. In this embodiment, the object feature information of the target object and the content feature information of the candidate resource are used to predict the situation of the interaction behavior of the target object and the candidate resource, so that the prediction accuracy of the behavior prediction parameter can be improved.

In some embodiments, the server employs a behavior prediction model to predict behavior prediction parameters of the candidate resource in the at least two behavior dimensions, corresponding to: for any behavior dimension, the server inputs the object feature information of the target object and the content feature information of the candidate resource into a behavior prediction model corresponding to the behavior dimension, and predicts the behavior prediction parameter of the candidate resource in the behavior dimension through the behavior prediction model to obtain the behavior prediction parameter of the candidate resource in the behavior dimension. In this way, the behavior prediction model corresponding to each behavior dimension is utilized to predict the interactive behavior of the target object and the candidate resource, so that not only is the prediction efficiency of the behavior prediction parameters improved, but also the prediction accuracy of the behavior prediction parameters is improved.

In an embodiment of the present disclosure, the behavior prediction model is a trained model. In some embodiments, for a behavior prediction model corresponding to any one of the behavior dimensions, the behavior prediction model is trained based on object feature information of a sample object, content feature information of candidate resources of the sample object, and behavior parameter labels of the candidate resources of the sample object in the behavior dimension. Wherein the behavior parameter tag represents an interaction behavior of the sample object with the candidate resource in the behavior dimension.

In some embodiments, for any one of the behavior dimensions, the server acquires object feature information of a plurality of sample objects, content feature information of candidate resources of the plurality of sample objects, and behavior parameter labels of the candidate resources of the plurality of sample objects in the behavior dimension, and performs model training based on the object feature information of the plurality of sample objects, the content feature information of the candidate resources of the plurality of sample objects, and the behavior parameter labels of the candidate resources of the plurality of sample objects in the behavior dimension, so as to obtain a behavior prediction model. Specifically, the training process of the behavior prediction model includes: in the first iteration process, respectively inputting object characteristic information of the plurality of sample objects and content characteristic information of candidate resources of the plurality of sample objects into an initial model to obtain a parameter training result of the first iteration process; determining a loss function based on a parameter training result of the first iteration process and a behavior parameter label of a corresponding sample object, and adjusting model parameters in an initial model based on the loss function; taking the model parameters after the first iteration adjustment as the model parameters of the second iteration, and then carrying out the second iteration; and repeating the iterative process for a plurality of times, and in the nth process, taking the model parameters after the N-1 th iterative adjustment as new model parameters, and performing model training until the training meets the target condition, and acquiring a model corresponding to the iterative process meeting the target condition as a behavior prediction model. Wherein N is a positive integer, and N is greater than 1. In some embodiments, the target condition satisfied by the training is that the number of training iterations of the initial model reaches a target number of training iterations, the target number being a preset number of training iterations; alternatively, the training may meet a target condition that the penalty value meets a target threshold condition, such as a penalty value less than 0.00001. The embodiments of the present disclosure do not limit the setting of the target condition.

In step 303, for any one of the behavioral dimensions, the server obtains a first parameter for the target object in the behavioral dimension, the first parameter representing a difference in the target object in the behavioral dimension relative to the population of objects.

Wherein, the object group refers to a user group. In some embodiments, the object group refers to an entire object using the application program with the multimedia resource playing function, or the object group refers to an object having registered the application program with the multimedia resource playing function.

In some embodiments, the server obtains a first average of the target object in the behavior dimension and a second average of the population of objects in the behavior dimension, and determines the first parameter based on the first average and the second average. The first average value is an average value of behavior prediction parameters of the candidate resources of the target object in the behavior dimension, and the second average value is an average value of behavior prediction parameters of the candidate resources of the plurality of objects included in the object group in the behavior dimension.

In an embodiment of the disclosure, the first parameter is positively correlated with the first average value, and the first parameter is negatively correlated with the second average value. Accordingly, the server determines the first parameter based on the first average value and the second average value by: in an alternative embodiment, the server determines a ratio of the first average value to the second average value, the ratio being determined as the first parameter; alternatively, in another alternative embodiment, a difference between the first average value and the second average value is determined, and the difference is determined as the first parameter. It should be noted that, the above manner of determining the ratio as the first parameter or determining the difference as the first parameter is an exemplary manner of the embodiments of the disclosure, and in other embodiments, the server may select other operation manners capable of representing the correlation between the first parameter and the first average value and the second average value to determine the first parameter, which is not limited in the embodiments of the disclosure. The embodiments of the present disclosure will be described with reference to determining a ratio as the first parameter.

In some embodiments, the server determines behavior prediction parameters for candidate resources of a plurality of objects included in the object population in respective behavior dimensions prior to implementing the solution. Optionally, the server periodically determines the behavior prediction parameters of the candidate resources of the plurality of objects included in the object group in each behavior dimension, and updates the behavior prediction parameters of the candidate resources of the plurality of objects included in the object group in each behavior dimension in real time.

For example, taking a target object as a target user, a candidate resource as a candidate video, and a behavior dimension as a praise dimension as an example, in the case that the number of candidate videos of the target user is 10, assuming that praise rates of the 10 candidate videos are a1, a2, … …, a9, and a10, respectively, a first average value is (a1+a2+ … … +a9+a10)/10; in the case where the number of candidate videos of a plurality of objects included in the object group is 100, assuming that the praise rates of the 100 candidate videos are b1, b2, … …, b99, b100, respectively, the second average value is (b1+b2+ … … +b99+b100)/100; further, the first parameter value is the ratio of (a1+a2+ … … +a9+a10)/10 to (b1+b2+ … … +b99+b100)/100.

In the above embodiment, the first parameter is determined based on the first average value and the second average value, and since the first average value is an average value of behavior prediction parameters of the target object in the behavior dimension, the second average value is an average value of behavior prediction parameters of a plurality of objects included in the object group in the behavior dimension, so that the determined first parameter can embody a difference of the target object in the behavior dimension relative to the object group, so that a weight coefficient corresponding to the behavior dimension is determined based on the first parameter.

In step 304, the server obtains a second parameter of the target object in the behavior dimension, where the second parameter represents a distribution of the plurality of candidate resources of the target object in the behavior dimension.

In some embodiments, the server obtains a dispersion of the behavior prediction parameters of the plurality of candidate resources of the target object in the behavior dimension, and determines the dispersion as the second parameter.

Wherein the dispersion is used to represent the degree of dispersion of one dataset. Accordingly, the process of determining the second parameter by the server is: in an alternative embodiment, the server obtains variances of the behavior prediction parameters of the plurality of candidate resources of the target object in the behavior dimension, and determines the variances as the second parameters; or, in another alternative embodiment, the server obtains standard deviations of the behavior prediction parameters of the plurality of candidate resources of the target object in the behavior dimension, and determines the standard deviations as the second parameters. It should be noted that, the above manner of determining the variance as the second parameter or determining the standard deviation as the second parameter is an exemplary manner of the embodiments of the disclosure, and in other embodiments, the server may select other parameters capable of representing the degree of discretization as the second parameter, which is not limited by the embodiments of the disclosure. The embodiments of the present disclosure will be described with reference to determining the standard deviation as the second parameter.

Illustratively, taking the standard deviation as the second parameter as an example, the standard deviation of the behavior prediction parameters of the plurality of candidate resources in the praise dimension can represent the distribution situation of the plurality of candidate resources in the praise dimension; the standard deviation of the behavior prediction parameters of the plurality of candidate resources in the collection dimension can represent the distribution situation of the plurality of candidate resources in the collection dimension.

In the above embodiment, the dispersion of the behavior prediction parameters of the plurality of candidate resources in the behavior dimension is determined as the second parameter, and since the dispersion can reflect the degree of dispersion of one data set, the determined second parameter can reflect the distribution situation of the plurality of candidate resources in the behavior dimension, so that the weight coefficient corresponding to the behavior dimension is determined based on the second parameter.

In step 305, the server determines a weight coefficient corresponding to the behavior dimension based on the first parameter and the second parameter of the target object in the behavior dimension.

In some embodiments, the server determines a third parameter of the target object in the behavioral dimension based on the first parameter and the second parameter of the target object in the behavioral dimension, wherein the third parameter is positively correlated with the first parameter and the third parameter is positively correlated with the second parameter; and determining a weight coefficient corresponding to the behavior dimension based on the initial weight coefficient corresponding to the behavior dimension and the third parameter, wherein the weight coefficient is positively correlated with the initial weight coefficient, and the weight coefficient is positively correlated with the third parameter.

In an embodiment of the disclosure, the third parameter is positively correlated with the first parameter, and the third parameter is positively correlated with the second parameter. The following describes a process of determining a third parameter by the server based on the first parameter and the second parameter, and the corresponding process is: determining a fourth parameter of the target object in the behavior dimension based on a second parameter of the target object in the behavior dimension, the fourth parameter being inversely related to the second parameter; the third parameter is determined based on the first parameter and the fourth parameter of the target object in the behavior dimension, the third parameter being inversely related to the fourth parameter.

Wherein the fourth parameter is inversely related to the second parameter. Accordingly, the process of determining the fourth parameter by the server based on the second parameter is: in an alternative embodiment, the server determines a difference from the second parameter based on the second parameter of the target object in the behavior dimension, and determines the difference as a fourth parameter of the target object in the behavior dimension; or, in another alternative embodiment, the server determines a square value of a difference value between the first parameter and the second parameter, and determines the square value as a fourth parameter of the target object in the behavior dimension. In this embodiment, the fourth parameter can be quickly determined by performing the foregoing process of taking the difference value or taking the square value of the difference value on the second parameter, and the determined fourth parameter can embody the negative correlation between the second parameter and the fourth parameter, so that the weight coefficient corresponding to the behavior dimension is determined based on the fourth parameter. Of course, in other embodiments, the server may alternatively use other operation manners capable of representing the negative correlation between the fourth parameter and the second parameter to determine the fourth parameter, for example, taking the reciprocal of the second parameter to obtain the fourth parameter, which is not limited by the embodiments of the present disclosure. The embodiment of the present disclosure will next describe a description of the embodiment taking the determination of the square value as an example of this fourth parameter.

Wherein the third parameter is positively correlated with the first parameter and the third parameter is negatively correlated with the fourth parameter. Correspondingly, the server determines the third parameter based on the first parameter and the fourth parameter as follows: in an alternative embodiment, the server determines a ratio of the first parameter to the fourth parameter, and determines the ratio as the third parameter; or, in another alternative embodiment, a difference between the first parameter and the fourth parameter is determined, and the difference is determined as the third parameter. It should be noted that, the above manner of determining the ratio as the third parameter or determining the difference as the third parameter is an exemplary manner of the embodiments of the disclosure, and in other embodiments, the server may select other operation manners capable of representing the correlation between the third parameter and the first parameter and the fourth parameter to determine the third parameter, which is not limited in the embodiments of the disclosure. The embodiment of the present disclosure will next take the determined ratio as an example of the third parameter to describe the description.

In the above embodiment, by determining an intermediate parameter (fourth parameter) that is inversely related to the second parameter, and then determining the third parameter based on the positive correlation between the third parameter and the first parameter and the negative correlation between the third parameter and the intermediate parameter, the determined third parameter can embody not only the positive correlation with the first parameter but also the positive correlation with the second parameter, so that the weight coefficient corresponding to the behavior dimension can be determined by using the third parameter later. It should be noted that, the foregoing manner of determining the fourth parameter first and then determining the third parameter is an exemplary manner of the embodiments of the disclosure, in other embodiments, the server may determine the third parameter by using other operation manners capable of representing a positive correlation between the third parameter and the first parameter, and the second parameter, for example, the server determines a sum of the first parameter and the second parameter as the third parameter; alternatively, the server determines the product of the first parameter and the second parameter as the third parameter, and so forth, to which embodiments of the present disclosure are not limited.

In an embodiment of the disclosure, the weight coefficient is positively correlated with the initial weight coefficient, and the weight coefficient is positively correlated with the third parameter. The following describes a process of determining the weight coefficient corresponding to the behavior dimension by the server based on the initial weight coefficient corresponding to the behavior dimension and the third parameter, and the corresponding process is as follows: in an alternative embodiment, the server determines a product of the initial weight coefficient corresponding to the behavior dimension and the third parameter, and determines the product as the weight coefficient corresponding to the behavior dimension; or in another optional embodiment, the server determines a sum of the initial weight coefficient corresponding to the behavior dimension and the third parameter, and determines the sum as the weight coefficient corresponding to the behavior dimension. It should be noted that, the above manner of determining the product as the weight coefficient or determining the sum as the weight coefficient is an exemplary manner of the embodiments of the disclosure, and in other embodiments, the server may select other operation manners capable of representing the correlation between the weight coefficient and the initial weight coefficient and the third parameter to determine the weight coefficient, which is not limited in the embodiments of the disclosure. The embodiments of the present disclosure will next describe the description of the embodiments taking the determination of the product as an example of the weight coefficient.

In some embodiments, for the above-mentioned processes from step 303 to step 305, the following optimization algorithm (1) may be used to determine the weight coefficient corresponding to each behavior dimension, where the corresponding process is: for any one of the behavior dimensions, the server determines a weight coefficient corresponding to the behavior dimension based on an initial weight coefficient corresponding to the behavior dimension, a first average value of the target object in the behavior dimension, a second average value of the object population in the behavior dimension, a second parameter of the target object in the behavior dimension, and the following optimization algorithm (1).

Wherein i represents a behavior dimension i, i is 0.ltoreq.i.ltoreq.K, wherein K represents the total number of the plurality of behavior dimensions, such as 5; u represents a target object; w (w) _i Representing an initial weight coefficient corresponding to the behavior dimension i, such as a weight coefficient set by service personnel according to historical experience;an optimization function representing the weight coefficient is used for outputting an optimized weight coefficient corresponding to the behavior dimension i; q _i,u Behavior prediction parameters representing multiple candidate resources of the target object u in the behavior dimension i, and accordingly avg (q _i,u ) An average value, i.e., a first average value, of the behavior prediction parameters of the plurality of candidate resources representing the target object u over the behavior dimension; all_user represents the object population; score _{i,all_user} Behavior prediction parameters representing candidate resources of a plurality of objects included in object population in behavior dimension i, 0.ltoreq.score _{i,all_user} Is less than or equal to 1, and accordingly, avg (score) _{i,all_user} ) An average value of the behavior prediction parameters, namely a second average value, of the candidate resources representing the plurality of objects included in the object group in the behavior dimension i; std (q) _i,u ) The dispersion of the behavior prediction parameters, i.e. the second parameters, of a plurality of the candidate resources representing the target object u in the behavior dimension i. The partial expression shown on the right side of the optimization algorithm (1) is the third parameter, and the denominator of the partial expression is the fourth parameter.

In the above embodiment, based on the positive correlation between the first parameter and the weight coefficient and the positive correlation between the second parameter and the weight coefficient, the third parameter can be quickly determined based on the first parameter and the second parameter, and further, by combining the initial weight coefficient corresponding to the behavior dimension, the weight coefficient corresponding to the behavior dimension can be quickly determined, and the accuracy of determining the weight coefficient is improved while the efficiency of determining the weight coefficient is improved.

For easy understanding, the optimization algorithm (1) is described in detail below: the partial expression shown on the right side of the optimization algorithm (1) is an optimization factor (or called a disturbance factor) of the weight coefficient, and the optimized weight coefficient can be obtained by multiplying the initial weight coefficient with the optimization factor. In the molecular part of the formula, the indicated significance factor of the object layer (i.e. the user layer) is specifically a ratio of a first average value corresponding to the target object (i.e. the user triggering the current resource push request) to a second average value corresponding to the object group (i.e. the whole users), and the ratio (i.e. the first parameter) can be used for measuring the difference between the target object and the object group. Thus, the tendency degree of the target object relative to the object group in the corresponding behavior dimension can be reflected through the ratio, for example, if the ratio is greater than 1, the average level of the target object relative to the object group is represented, the tendency (such as praise, attention, etc.) in the corresponding behavior dimension is strong, and the weight coefficient of the behavior dimension can be properly increased at this time; if the ratio is smaller than 1, the average level of the target object relative to the object group is indicated, and the target object has weaker tendency in the corresponding behavior dimension, and the weight coefficient corresponding to the behavior dimension can be properly reduced. In the denominator part of the partial formula, indicated is a significance factor at the candidate resource level, in particular a dispersion, such as a standard deviation, of the behavior prediction parameters of the plurality of candidate resources of the target object in the behavior dimension. Thus, through the dispersion (i.e., the second parameter), the distribution situation of the plurality of candidate resources in the behavior dimension can be represented, for example, taking the standard deviation as an example, if the standard deviation is larger, the distribution of the plurality of candidate resources in the behavior dimension is represented to be more discrete, and at the moment, the candidate resources have more obvious differentiation degree, so that larger contribution can be generated for subsequent resource pushing, and the weight coefficient corresponding to the behavior dimension can be appropriately increased; if the standard deviation is smaller, the distribution of the plurality of candidate resources on the behavior dimension is concentrated, and the candidate resources do not have obvious distinction, so that the weight coefficient corresponding to the behavior dimension can be properly reduced.

It should be understood that, for each object-triggered resource push request, the push result of the resource push request may be represented by f (u, r), where u represents the current object (e.g., the target object), r represents the resource push request triggered by the current object, and f (u, r) represents that the push result of the resource push request is related to the current object and the resource push request triggered by the current object. In the face of a multi-objective optimization problem for resource pushing (i.e., corresponding to multiple behavior dimensions), a request may be embodied as a combination of multiple behavior dimensions, and accordingly, the pushing result of the resource pushing request may be represented as Φ (f (u, q _i,u )│w _i ) Where Φ is a fusion function of multiple behavior dimensions, q _i,u Behavior prediction parameters, w, representing candidate resources of current object u in behavior dimension i _i And representing the initial weight coefficient corresponding to the behavior dimension i. It can be seen that, for a single behavior dimension, under the condition of determining the initial weight coefficient, the push result corresponding to the behavior dimension is matched with the object u and the candidate resource q _i,u In other words, the difference in the push result of one request is mainly expressed in both the difference in the subject individual and the difference in the candidate resource.

Taking the like dimension as an example, in the case that the user a and the user B trigger a resource push request respectively, the server respectively screens n candidate videos for the user a and the user B through a recall stage and a coarse ranking stage, and assuming that the user a has more frequent like operations relative to the user B, the like rate predicted for the n candidate videos of the user a is generally higher than the like rate predicted for the n candidate videos of the user B, which is the difference between the individual users. In addition, for n candidate videos of the same user, there is a difference, and assuming that the praise rates of the n candidate videos of the user a are all 0.5, the praise rate of the candidate videos of the user B is distributed between 0 and 1, it can be found that the praise rate distribution of the user a is concentrated, and the praise rate distribution of the user B is discrete, that is, the difference between the candidate videos.

In the embodiment of the disclosure, based on the two aspects of the difference of the subject individual and the difference of the candidate resource, an optimization method of the weight coefficient is provided, so that the weight coefficient of each behavior dimension is not fixed, but the weight coefficient corresponding to each behavior dimension can be adaptively adjusted based on the difference of the subject individual and the difference of the candidate resource, thereby realizing personalized adjustment of the weight coefficient and personalized pushing of the multimedia resource. At this time, the push result of one request may be expressed as Wherein->I.e. the optimization function shown in the optimization algorithm (1) above.

For the above procedure of determining the second parameter, taking the determination of the standard deviation as the second parameter as an example, in some embodiments, the server uses the following operation formula (2) to determine the standard deviation of the behavior prediction parameters of the plurality of candidate resources of the target object in the behavior dimension.

Wherein x represents a random variable corresponding to the behavior prediction parameters of each candidate resource mentioned in the above embodiments; e (x) represents an average value of behavior prediction parameters of a plurality of candidate resources; x is x ² Also representing random variables; e (x) ² ) An average value representing square values of behavior prediction parameters of a plurality of candidate resources; incidentally, E (x ² )-E(x) ² Representing variances of behavior prediction parameters of the plurality of candidate resources in the behavior dimension. Thus, the average value and the standard deviation can be calculated simultaneously in one operation process by the operation formula (2), thusThe complexity of the operation formula (2) is O (n), where O (n) refers to the complexity of the operation formula (also referred to as linear time), and is used to represent that the complexity of the operation formula (2) increases linearly and is proportional to the input number. For the multi-objective optimization problem of K behavior dimensions related to the embodiments of the present disclosure, the overall complexity of the embodiments of the present disclosure may be represented as O (nK), while considering that the number of candidate resources in the multi-objective optimization problem is typically constant (n.ltoreq.500), the complexity of the resource recommendation method provided by the present disclosure is O (K). Therefore, the operation complexity of the resource pushing method is reduced, and the operation efficiency of the resource pushing method is improved.

In the above embodiment, a manner of optimizing the initial weight coefficient based on the first parameter and the second parameter to obtain the optimized weight coefficient is provided, so that the determined weight coefficient can embody a positive correlation with the first parameter and simultaneously embody a positive correlation with the second parameter. It should be noted that, the optimization manner of the weight coefficient shown above is an exemplary manner of the embodiments of the disclosure, in other embodiments, the server may select other operation manners capable of representing the positive correlation between the weight coefficient and the first parameter, and the second parameter to optimize the initial weight coefficient, for example, the server determines the product of the initial weight coefficient corresponding to the behavior dimension, the first parameter and the second parameter, determines the product as the optimized weight coefficient, and so on, which is not limited in the embodiments of the disclosure.

In step 306, the server determines a pushing parameter of the candidate resource based on the weight coefficients of the behavior prediction parameter of the candidate resource in the at least two behavior dimensions and the at least two behavior dimensions, where the pushing parameter is used to determine a pushing order of the candidate resource when pushing the resource for the target object.

The server determines the pushing parameters of the candidate resource based on the weight coefficients of the behavior prediction parameters of the candidate resource in the at least two behavior dimensions and the corresponding at least two behavior dimensions, and the process refers to steps (306A) to (306C):

(306A) The server sorts the candidate resources in the at least two behavior dimensions according to the behavior prediction parameters of the candidate resources in the at least two behavior dimensions.

In some embodiments, the server sorts the candidate resources according to the behavior prediction parameters of the candidate resources in the queues corresponding to the at least two behavior dimensions.

In an alternative embodiment, when the server ranks the behavior prediction parameters in any one behavior dimension, responding that the behavior dimension belongs to a first behavior dimension, and ranking according to the ranking order of the behavior prediction parameters of the candidate resource in the behavior dimension from high to low; in other embodiments, responsive to the behavior dimension belonging to the second behavior dimension, the candidate resource is ranked in order of low-to-high ranking of the behavior prediction parameters in the behavior dimension. The first behavior dimension includes a positive behavior dimension and a viewing duration dimension of the multimedia resource, wherein the positive behavior dimension is a behavior dimension corresponding to a positive interaction behavior (positive interaction) such as a clicking behavior, a collecting behavior, and the like, and the second behavior dimension represents a negative behavior dimension corresponding to a negative interaction (negative interaction) such as a shielding behavior, a skipping behavior, and the like.

For example, taking at least two behavior dimensions as a praise dimension, a mask dimension and a viewing duration dimension of a multimedia resource as examples, the queues corresponding to the at least two behavior dimensions may be a praise queue, a mask queue and a viewing duration queue of the multimedia resource, further, the praise queues are ordered according to the order of the praise rate of the candidate resource from high to low, the mask queues are ordered according to the order of the mask rate of the candidate resource from low to high, and the viewing duration queues of the multimedia resource are ordered according to the order of the predicted viewing duration of the candidate resource from high to low.

In the above embodiment, the sorting process of the behavior prediction parameters on each behavior dimension is performed by setting the queue corresponding to each behavior dimension so that the subsequent sorting process is performed on the queue corresponding to each behavior dimension.

(306B) The server determines push sub-parameters of the candidate resource in the at least two behavior dimensions based on the arrangement order of the candidate resource in the at least two behavior dimensions, respectively.

The pushing subparameter represents the ordering condition of the candidate resources in the corresponding behavior dimension. Taking a queue as an example, the push subparameter represents the ordering condition of the candidate resources in the corresponding queue.

In the disclosed embodiment, the push subparameter is inversely related to the arrangement order. Correspondingly, the process of determining the pushing subparameter by the server is as follows: in an alternative embodiment, the server determines the reciprocal of the order of arrangement of the candidate resource in the at least two behavior dimensions, and determines the reciprocal as a push subparameter of the candidate resource in the at least two behavior dimensions; or, in another alternative embodiment, the server determines a difference from the ranking order based on the ranking order of the candidate resource in the at least two behavior dimensions, and determines the difference as a push subparameter of the candidate resource in the at least two behavior dimensions. In the embodiment, the pushing sub-parameters of the candidate resources in each behavior dimension can be rapidly determined by taking the reciprocal or the difference value of the arrangement order of the candidate resources in each behavior dimension, so that the efficiency of determining the pushing sub-parameters is improved.

(306C) The server determines the pushing parameters of the candidate resource based on the pushing sub-parameters of the candidate resource in the at least two behavior dimensions and the weight coefficients corresponding to the at least two behavior dimensions.

In some embodiments, the server performs weighted summation based on the pushing sub-parameters of the candidate resource in the at least two behavior dimensions and the weight coefficients corresponding to the at least two behavior dimensions, so as to obtain the pushing parameters of the candidate resource. In the embodiment, the fusion of the behavior prediction parameters in each behavior dimension is realized in a weighted summation mode, so that the pushing parameters with high accuracy can be determined, and further, the resource pushing is performed based on the pushing parameters, thereby improving the efficiency and accuracy of the resource pushing.

In the above embodiment, based on the arrangement order of the candidate resources in each behavior dimension, the pushing sub-parameters of the candidate resources in each behavior dimension are determined, so that the determined pushing sub-parameters can embody the ordering condition of the candidate resources, and further, based on the pushing sub-parameters of the candidate resources in each behavior dimension and the weight coefficients corresponding to each behavior dimension, the pushing parameters of the candidate resources are determined, so that the pushing sub-parameters of the candidate resources in a plurality of behavior dimensions are synthesized, the pushing parameters with high accuracy can be determined, and further, the resource pushing is performed based on the pushing parameters, thereby improving the efficiency and accuracy of the resource pushing.

In step 307, the server pushes the resource for the target object based on the push parameters of the candidate resource.

In some embodiments, the pushing parameters of the plurality of candidate resources of the target object can be obtained through steps 301 to 306, and then the plurality of candidate resources are sorted according to the ranking order of the plurality of candidate resources from high to low, and then the multimedia resources corresponding to the ranking order are sequentially pushed to the target object according to the ranking order of the plurality of candidate resources.

For example, fig. 4 is a flowchart of a resource pushing method according to an exemplary embodiment, referring to fig. 4, first, predicting n candidate videos (i.e., candidate resources) of a target user (i.e., a target object) by using a behavior prediction model, respectively obtaining behavior prediction values (i.e., behavior prediction parameters) in K behavior dimensions, and generating K behavior queues according to the behavior prediction values of the n candidate videos in the K behavior dimensions; calculating disturbance factors corresponding to each behavior dimension by using the optimization algorithm (1), optimizing initial weight coefficients of each behavior dimension to obtain optimized weight coefficients, particularly, under the condition of confidence of a behavior prediction model, calculating average values of behavior predicted values of n candidate videos of a target user and making a ratio with the average values of the behavior predicted values of a plurality of users included in a user group (namely, an object group) to reflect the tendency degree of the target user in the current behavior dimension, and calculating standard deviation of the behavior predicted values of n candidate videos of the target user in the current behavior dimension to reflect the distribution situation of a plurality of candidate videos of the target user in the current behavior dimension, and optimizing the initial weight coefficients of each behavior dimension by using the optimization algorithm (1) to obtain optimized weight coefficients; and determining the ranking score (the enstable score, namely the pushing parameter) of the candidate videos by utilizing the internal scores (the rank score, namely the pushing subparameter) of the candidate videos in the queues corresponding to the behavior dimensions and the optimized weight coefficient corresponding to the behavior dimensions, acquiring the ranking scores of the plurality of candidate videos based on the process, ranking the plurality of candidate videos based on the ranking scores of the plurality of candidate videos, and pushing the resources according to the ranking order of the plurality of candidate resources.

It should be noted that, in the fine-ranking stage of resource pushing, the resource pushing method provided by the embodiment of the present disclosure is applied, so that on one hand, the use duration of the application program and the user retention rate of the application program are improved, and on the other hand, the effective playing times of the multimedia resource and the interaction rate based on the multimedia resource are also improved.

Fig. 5 is a block diagram illustrating a resource pushing device according to an example embodiment. Referring to fig. 5, the apparatus includes an acquisition unit 501, a weight coefficient determination unit 502, and a push parameter determination unit 503.

An obtaining unit 501 configured to perform obtaining a behavior prediction parameter of a candidate resource of a target object in at least two behavior dimensions, where the behavior prediction parameter represents a predicted situation of an interaction behavior of the target object with the candidate resource in the behavior dimension;

a weight coefficient determining unit 502 configured to perform determining a weight coefficient corresponding to the behavior dimension based on a first parameter and a second parameter of the target object in the behavior dimension, the first parameter representing a difference of the target object in the behavior dimension relative to a population of objects, the second parameter representing a distribution of the plurality of candidate resources of the target object in the behavior dimension;

A pushing parameter determining unit 503 configured to determine a pushing parameter of the candidate resource based on the weight coefficients of the behavior prediction parameter of the candidate resource in the at least two behavior dimensions and the at least two behavior dimensions, where the pushing parameter is used to determine a pushing order of the candidate resource when pushing the resource for the target object.

In some embodiments, the weight coefficient determination unit 502 includes:

In some embodiments, the obtaining unit 501 is configured to perform:

In some embodiments, the push parameter determining unit 503 includes:

In some embodiments, the ordering subunit is configured to perform:

In some embodiments, the determining subunit is further configured to perform:

It should be noted that: in the resource pushing device provided in the above embodiment, only the division of the above functional modules is used for illustration, and in practical application, the above functional allocation may be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the resource pushing device and the resource pushing method provided in the foregoing embodiments belong to the same concept, and specific implementation processes of the resource pushing device and the resource pushing method are detailed in the method embodiments and are not repeated herein.

The computer device mentioned in the embodiments of the present disclosure may be provided as a terminal. Fig. 6 shows a block diagram of a terminal 600 provided by an exemplary embodiment of the present disclosure. The terminal 600 may be: a smart phone, a tablet computer, an MP3 player (Moving Picture Experts Group Audio Layer III, motion picture expert compression standard audio plane 3), an MP4 (Moving Picture Experts Group Audio Layer IV, motion picture expert compression standard audio plane 4) player, a notebook computer, or a desktop computer. Terminal 600 may also be referred to by other names of user devices, portable terminals, laptop terminals, desktop terminals, etc.

In general, the terminal 600 includes: a processor 601 and a memory 602.

Processor 601 may include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. The processor 601 may be implemented in at least one hardware form of DSP (Digital Signal Processing ), FPGA (Field-Progra mmable Gate Array, field programmable gate array), PLA (Programmable Logic Array ). The processor 601 may also include a main processor, which is a processor for processing data in an awake state, also called a CPU (Central Processing Unit ), and a coprocessor; a coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 601 may be integrated with a GPU (Graphics Processing Unit, image processor) for taking care of rendering and rendering of content that the display screen is required to display. In some embodiments, the processor 601 may also include an AI (Artificial Intelligence ) processor for processing computing operations related to machine learning.

The memory 602 may include one or more computer-readable storage media, which may be non-transitory. The memory 602 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In some embodiments, a non-transitory computer readable storage medium in memory 602 is used to store at least one program code for execution by processor 601 to implement the processes performed by a terminal in the resource pushing method provided by the method embodiments in the present disclosure.

In some embodiments, the terminal 600 may further optionally include: a peripheral interface 603, and at least one peripheral. The processor 601, memory 602, and peripheral interface 603 may be connected by a bus or signal line. The individual peripheral devices may be connected to the peripheral device interface 603 via buses, signal lines or a circuit board. Specifically, the peripheral device includes: at least one of radio frequency circuitry 604, a display 605, a camera assembly 606, audio circuitry 607, a positioning assembly 608, and a power supply 609.

Peripheral interface 603 may be used to connect at least one Input/Output (I/O) related peripheral to processor 601 and memory 602. In some embodiments, the processor 601, memory 602, and peripheral interface 603 are integrated on the same chip or circuit board; in some other embodiments, either or both of the processor 601, memory 602, and peripheral interface 603 may be implemented on separate chips or circuit boards, which is not limited in this embodiment.

The Radio Frequency circuit 604 is configured to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 604 communicates with a communication network and other communication devices via electromagnetic signals. The radio frequency circuit 604 converts an electrical signal into an electromagnetic signal for transmission, or converts a received electromagnetic signal into an electrical signal. Optionally, the radio frequency circuit 604 includes: antenna systems, RF transceivers, one or more amplifiers, tuners, oscillators, digital signal processors, codec chipsets, subscriber identity module cards, and so forth. The radio frequency circuit 604 may communicate with other terminals via at least one wireless communication protocol. The wireless communication protocol includes, but is not limited to: metropolitan area networks, various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks, and/or WiFi (Wireless Fidelity ) networks. In some embodiments, the radio frequency circuitry 604 may also include NFC (Near Field Communication, short range wireless communication) related circuitry, which is not limited by the present disclosure.

The display screen 605 is used to display a UI (User Interface). The UI may include graphics, text, icons, video, and any combination thereof. When the display 605 is a touch display, the display 605 also has the ability to collect touch signals at or above the surface of the display 605. The touch signal may be input as a control signal to the processor 601 for processing. At this point, the display 605 may also be used to provide virtual buttons and/or virtual keyboards, also referred to as soft buttons and/or soft keyboards. In some embodiments, the display 605 may be one, disposed on the front panel of the terminal 600; in other embodiments, the display 605 may be at least two, respectively disposed on different surfaces of the terminal 600 or in a folded design; in other embodiments, the display 605 may be a flexible display, disposed on a curved surface or a folded surface of the terminal 600. Even more, the display 605 may be arranged in a non-rectangular irregular pattern, i.e., a shaped screen. The display 605 may be made of LCD (Liquid Crystal Display ), OLED (Organic Light-Emitting Diode) or other materials.

The camera assembly 606 is used to capture images or video. Optionally, the camera assembly 606 includes a front camera and a rear camera. Typically, the front camera is disposed on the front panel of the terminal and the rear camera is disposed on the rear surface of the terminal. In some embodiments, the at least two rear cameras are any one of a main camera, a depth camera, a wide-angle camera and a tele camera, so as to realize that the main camera and the depth camera are fused to realize a background blurring function, and the main camera and the wide-angle camera are fused to realize a panoramic shooting and Virtual Reality (VR) shooting function or other fusion shooting functions. In some embodiments, camera assembly 606 may also include a flash. The flash lamp can be a single-color temperature flash lamp or a double-color temperature flash lamp. The dual-color temperature flash lamp refers to a combination of a warm light flash lamp and a cold light flash lamp, and can be used for light compensation under different color temperatures.

The audio circuit 607 may include a microphone and a speaker. The microphone is used for collecting sound waves of users and environments, converting the sound waves into electric signals, and inputting the electric signals to the processor 601 for processing, or inputting the electric signals to the radio frequency circuit 604 for voice communication. For the purpose of stereo acquisition or noise reduction, a plurality of microphones may be respectively disposed at different portions of the terminal 600. The microphone may also be an array microphone or an omni-directional pickup microphone. The speaker is used to convert electrical signals from the processor 601 or the radio frequency circuit 604 into sound waves. The speaker may be a conventional thin film speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, not only the electric signal can be converted into a sound wave audible to humans, but also the electric signal can be converted into a sound wave inaudible to humans for ranging and other purposes. In some embodiments, the audio circuit 607 may also include a headphone jack.

The location component 608 is used to locate the current geographic location of the terminal 600 to enable navigation or LBS (Location Based Service, location based services).

A power supply 609 is used to power the various components in the terminal 600. The power source 609 may be alternating current, direct current, disposable battery or rechargeable battery. When the power source 609 includes a rechargeable battery, the rechargeable battery may support wired or wireless charging. The rechargeable battery may also be used to support fast charge technology.

In some embodiments, the terminal 600 further includes one or more sensors 610. The one or more sensors 610 include, but are not limited to: acceleration sensor 611, gyroscope sensor 612, pressure sensor 613, fingerprint sensor 614, optical sensor 615, and proximity sensor 616.

The acceleration sensor 611 can detect the magnitudes of accelerations on three coordinate axes of the coordinate system established with the terminal 600. For example, the acceleration sensor 611 may be used to detect components of gravitational acceleration in three coordinate axes. The processor 601 may control the display screen 605 to display the user interface in a landscape view or a portrait view according to the gravitational acceleration signal acquired by the acceleration sensor 611. The acceleration sensor 611 may also be used for the acquisition of motion data of a game or a user.

The gyro sensor 612 may detect a body direction and a rotation angle of the terminal 600, and the gyro sensor 612 may collect a 3D motion of the user on the terminal 600 in cooperation with the acceleration sensor 611. The processor 601 may implement the following functions based on the data collected by the gyro sensor 612: motion sensing (e.g., changing UI according to a tilting operation by a user), image stabilization at shooting, game control, and inertial navigation.

The pressure sensor 613 may be disposed at a side frame of the terminal 600 and/or at a lower layer of the display 605. When the pressure sensor 613 is disposed at a side frame of the terminal 600, a grip signal of the terminal 600 by a user may be detected, and a left-right hand recognition or a shortcut operation may be performed by the processor 601 according to the grip signal collected by the pressure sensor 613. When the pressure sensor 613 is disposed at the lower layer of the display screen 605, the processor 601 controls the operability control on the UI interface according to the pressure operation of the user on the display screen 605. The operability controls include at least one of a button control, a scroll bar control, an icon control, and a menu control.

The fingerprint sensor 614 is used for collecting the fingerprint of the user, and the processor 601 identifies the identity of the user according to the fingerprint collected by the fingerprint sensor 614, or the fingerprint sensor 614 identifies the identity of the user according to the collected fingerprint. Upon recognizing that the user's identity is a trusted identity, the processor 601 authorizes the user to perform relevant sensitive operations including unlocking the screen, viewing encrypted information, downloading software, paying for and changing settings, etc. The fingerprint sensor 614 may be disposed on the front, back, or side of the terminal 600. When a physical key or vendor Logo is provided on the terminal 600, the fingerprint sensor 614 may be integrated with the physical key or vendor Logo.

The optical sensor 615 is used to collect ambient light intensity. In one embodiment, processor 601 may control the display brightness of display 605 based on the intensity of ambient light collected by optical sensor 615. Specifically, when the intensity of the ambient light is high, the display brightness of the display screen 605 is turned up; when the ambient light intensity is low, the display brightness of the display screen 605 is turned down. In another embodiment, the processor 601 may also dynamically adjust the shooting parameters of the camera assembly 606 based on the ambient light intensity collected by the optical sensor 615.

A proximity sensor 616, also referred to as a distance sensor, is typically provided on the front panel of the terminal 600. The proximity sensor 616 is used to collect the distance between the user and the front of the terminal 600. In one embodiment, when the proximity sensor 616 detects a gradual decrease in the distance between the user and the front face of the terminal 600, the processor 601 controls the display 605 to switch from the bright screen state to the off screen state; when the proximity sensor 616 detects that the distance between the user and the front surface of the terminal 600 gradually increases, the processor 601 controls the display screen 605 to switch from the off-screen state to the on-screen state.

Those skilled in the art will appreciate that the structure shown in fig. 6 is not limiting of the terminal 600 and may include more or fewer components than shown, or may combine certain components, or may employ a different arrangement of components.

The computer device mentioned in the embodiments of the present disclosure may be provided as a server. Fig. 7 is a block diagram of a server according to an exemplary embodiment, where the server 700 may have a relatively large difference due to configuration or performance, and may include one or more processors (Central Processing Units, CPU) 701 and one or more memories 702, where the one or more memories 702 store at least one program code, and the at least one program code is loaded and executed by the one or more processors 701 to implement the processes performed by the server in the resource pushing method provided in the above embodiments. Of course, the server 700 may also have a wired or wireless network interface, a keyboard, an input/output interface, and other components for implementing the functions of the device, which are not described herein.

In an exemplary embodiment, a computer readable storage medium is also provided, e.g. a memory 702 comprising program code, which is executable by the processor 701 of the server 700 to perform the above described resource pushing method. Alternatively, the computer readable storage medium may be a ROM (Read-Only Memory), a RAM (Random Access Memory ), a CD-ROM (Compact-Disc Read-Only Memory), a magnetic tape, a floppy disk, an optical data storage device, and the like.

In an exemplary embodiment, a computer program product is also provided, comprising a computer program which, when executed by a processor, implements the above-mentioned resource pushing method.

In some embodiments, the computer program related to the embodiments of the present disclosure may be deployed to be executed on one computer device or on multiple computer devices located at one site, or alternatively, may be executed on multiple computer devices distributed across multiple sites and interconnected by a communication network, where the multiple computer devices distributed across multiple sites and interconnected by a communication network may constitute a blockchain system.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This disclosure is intended to cover any adaptations, uses, or adaptations of the disclosure following the general principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for pushing resources, the method comprising:

determining pushing parameters of the candidate resources based on the behavior prediction parameters of the candidate resources in the at least two behavior dimensions and weight coefficients corresponding to the at least two behavior dimensions, wherein the pushing parameters are used for determining pushing orders of the candidate resources when the candidate resources are pushed for the target object.

2. The resource pushing method according to claim 1, wherein the acquiring process of the first parameter includes:

determining the first parameter based on the first average value and the second average value, wherein the first parameter is positively correlated with the first average value, and the first parameter is negatively correlated with the second average value.

3. The resource pushing method according to claim 1, wherein the obtaining process of the second parameter includes:

4. The resource pushing method according to claim 1, wherein the determining the weight coefficient corresponding to the behavior dimension based on the first parameter and the second parameter of the target object in the behavior dimension includes:

5. The resource pushing method of claim 4, wherein the determining a third parameter of the target object in the behavior dimension based on the first parameter and the second parameter of the target object in the behavior dimension comprises:

the third parameter is determined based on the first parameter and a fourth parameter of the target object in the behavior dimension, and the third parameter is inversely related to the fourth parameter.

6. The resource pushing method according to claim 1, wherein the obtaining behavior prediction parameters of the candidate resource of the target object in at least two behavior dimensions includes:

and determining the behavior prediction parameters of the candidate resource in the at least two behavior dimensions based on the object characteristic information of the target object and the content characteristic information of the candidate resource.

7. A resource pushing device, the device comprising:

an obtaining unit configured to perform obtaining a behavior prediction parameter of a candidate resource of a target object in at least two behavior dimensions, the behavior prediction parameter representing a predicted situation in which the target object interacts with the candidate resource in the behavior dimensions;

a weight coefficient determination unit configured to perform determination of a weight coefficient corresponding to the behavior dimension based on a first parameter and a second parameter of the target object in the behavior dimension, the first parameter representing a difference of the target object in the behavior dimension relative to a population of objects, the second parameter representing a distribution of a plurality of the candidate resources of the target object in the behavior dimension;

And a pushing parameter determining unit configured to determine a pushing parameter of the candidate resource based on a weight coefficient of a behavior prediction parameter of the candidate resource in the at least two behavior dimensions and corresponding to the at least two behavior dimensions, where the pushing parameter is used to determine a pushing order of the candidate resource when pushing the resource for the target object.

8. A computer device, the computer device comprising:

one or more processors;

a memory for storing the processor-executable program code;

wherein the processor is configured to execute the program code to implement the resource pushing method of any of claims 1 to 6.

9. A computer readable storage medium, characterized in that program code in the computer readable storage medium, when executed by a processor of a computer device, enables the computer device to perform the resource pushing method of any of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the resource pushing method of any of claims 1 to 6.