CN112449002A

CN112449002A - Method, device and equipment for pushing object to be pushed and storage medium

Info

Publication number: CN112449002A
Application number: CN202011119839.9A
Authority: CN
Inventors: 马中团; 洪庚伟; 李羽
Original assignee: Weimin Insurance Agency Co Ltd
Current assignee: Weimin Insurance Agency Co Ltd
Priority date: 2020-10-19
Filing date: 2020-10-19
Publication date: 2021-03-05
Anticipated expiration: 2040-10-19
Also published as: CN112449002B

Abstract

The application relates to a pushing method, a pushing device, equipment and a storage medium of an object to be pushed, wherein the method comprises the following steps: acquiring a first object set from objects to be pushed corresponding to a target account, wherein the first object set comprises a plurality of objects with mutually exclusive relationship; screening a plurality of objects according to the target account characteristics of the target account to obtain a second object set, wherein the second object set comprises one target object matched with the target account characteristics in the plurality of objects, or the second object set is an empty set; replacing a first object set in the objects to be pushed with a second object set to obtain a target pushing object; and pushing the target pushing object to the target account. The technical problem that effective pushing cannot be performed on mutually exclusive products in the related technology is solved.

Description

Method, device and equipment for pushing object to be pushed and storage medium

Technical Field

The present application relates to the field of computers, and in particular, to a method, an apparatus, a device, and a storage medium for pushing an object to be pushed.

Background

With the popularization of information technology and the internet, intelligent recommendation is more and more commonly applied in various fields, and related technologies are rapidly developed and matured all the time. Conventional recommendation techniques focus on finding items that may be of interest to a user from a large number of items based on user behavior data. In recent years, with the development of big data related technologies, people also want to fully apply intelligent recommendation technologies in the vertical field, so as to show more suitable products for users and realize accurate marketing.

However, the conventional recommendation technology has some defects in the vertical field, and the current more mature recommendation technology mainly includes the following: content-based recommendations, demographic-based recommendations, collaborative-filtering-based recommendations, association-rule-based recommendations, other recommendation algorithms, and the like. These algorithms are not applicable to specific problems, such as mutually exclusive item recommendations. The traditional recommendation methods mainly solve the recommendation problem of similar articles, and are not suitable for distinguishing and recommending mutually exclusive articles.

In view of the above problems, no effective solution has been proposed.

Disclosure of Invention

The application provides a pushing method, a pushing device, pushing equipment and a storage medium of an object to be pushed, so as to at least solve the technical problem that effective pushing cannot be performed on mutually exclusive products in the related technology.

According to an aspect of the embodiments of the present application, a method for pushing an object to be pushed is provided, including:

acquiring a first object set from objects to be pushed corresponding to a target account, wherein the first object set comprises a plurality of objects with mutually exclusive relationship;

screening the plurality of objects according to the target account characteristics of the target account to obtain a second object set, wherein the second object set comprises one target object matched with the target account characteristics in the plurality of objects, or the second object set is an empty set;

replacing the first object set in the object to be pushed with the second object set to obtain a target pushing object;

and pushing the target pushing object to the target account.

According to another aspect of the embodiments of the present application, there is also provided a pushing apparatus for an object to be pushed, including:

the system comprises an acquisition module, a pushing module and a pushing module, wherein the acquisition module is used for acquiring a first object set from objects to be pushed corresponding to a target account, and the first object set comprises a plurality of objects which have mutual exclusion relationship;

the screening module is configured to screen the multiple objects according to the target account characteristics of the target account to obtain a second object set, where the second object set includes one of the multiple objects that matches the target account characteristics, or the second object set is an empty set;

the replacing module is used for replacing the first object set in the object to be pushed with the second object set to obtain a target pushing object;

and the pushing module is used for pushing the target pushing object to the target account.

According to another aspect of the embodiments of the present application, there is also provided a storage medium including a stored program which, when executed, performs the above-described method.

According to another aspect of the embodiments of the present application, there is also provided an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the above method through the computer program.

According to another aspect of embodiments of the present application, there is also provided a computer program product or a computer program comprising computer instructions stored in a computer readable storage medium; the processor of the computer device reads the computer instruction from the computer-readable storage medium, and executes the computer instruction, so that the computer device executes the steps of any embodiment of the pushing method for the object to be pushed.

In the embodiment of the application, a first object set is obtained from objects to be pushed corresponding to a target account, wherein the first object set comprises a plurality of objects which have mutual exclusive relationship; screening a plurality of objects according to the target account characteristics of the target account to obtain a second object set, wherein the second object set comprises one target object matched with the target account characteristics in the plurality of objects, or the second object set is an empty set; replacing a first object set in the objects to be pushed with a second object set to obtain a target pushing object; a mode of pushing a target pushing object to a target account is that for an object to be pushed corresponding to the target account, a first object set comprising a plurality of objects with mutual exclusive relationship is obtained, the plurality of objects are screened according to the characteristics of the target account, one target object is screened out or none of the plurality of objects is selected, a second object set is obtained, the second object set is used for replacing the first object set in the object to be pushed, so that the objects with mutual exclusive relationship in the objects to be pushed are only one of the target accounts or not pushed, thereby not only ensuring that a certain class of objects are recommended to proper users, but also distinguishing and distinguishing different user groups suitable for the objects with mutual exclusive relationship, therefore, the exposure utilization rate is improved, the commercial value of the display list is optimized, and effective pushing is carried out on mutually exclusive products.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive exercise.

Fig. 1 is a schematic diagram of a hardware environment of a pushing method of an object to be pushed according to an embodiment of the present application;

fig. 2 is a flowchart of an optional push method for an object to be pushed according to an embodiment of the present application;

FIG. 3 is a schematic diagram of an adaptive bucket partitioning mechanism according to an embodiment of the present application;

FIG. 4 is a schematic diagram of a system architecture of a product recommendation system in accordance with an alternative embodiment of the present application;

fig. 5 is a schematic diagram of an alternative pushing apparatus for pushing an object to be pushed according to an embodiment of the present application;

fig. 6 is a block diagram of an electronic device according to an embodiment of the present application.

Detailed Description

In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only partial embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

According to an aspect of the embodiments of the present application, an embodiment of a method for pushing an object to be pushed is provided.

Alternatively, in this embodiment, the pushing method of the object to be pushed may be applied to a hardware environment formed by the terminal 101 and the server 103 as shown in fig. 1. As shown in fig. 1, a server 103 is connected to a terminal 101 through a network, which may be used to provide services (such as game services, application services, etc.) for the terminal or a client installed on the terminal, and a database may be provided on the server or separately from the server for providing data storage services for the server 103, and the network includes but is not limited to: the terminal 101 is not limited to a PC, a mobile phone, a tablet computer, and the like. The pushing method of the object to be pushed according to the embodiment of the present application may be executed by the server 103, or may be executed by the terminal 101, or may be executed by both the server 103 and the terminal 101. The terminal 101 executing the method for pushing the object to be pushed according to the embodiment of the present application may also be executed by a client installed thereon.

Fig. 2 is a flowchart of an optional push method for an object to be pushed according to an embodiment of the present application, and as shown in fig. 2, the method may include the following steps:

step S202, a first object set is obtained from objects to be pushed corresponding to a target account, wherein the first object set comprises a plurality of objects which mutually have a mutual exclusion relationship;

step S204, screening the plurality of objects according to the target account characteristics of the target account to obtain a second object set, wherein the second object set comprises one target object matched with the target account characteristics in the plurality of objects, or the second object set is an empty set;

step S206, replacing the first object set in the object to be pushed with the second object set to obtain a target pushing object;

step S208, pushing the target push object to the target account.

Optionally, in this embodiment, the pushing method of the object to be pushed may be applied to, but not limited to, scenes in which products are intelligently recommended for users in various types of applications. The various types of applications described above may include, but are not limited to: financial applications, insurance applications, medical applications, educational applications, gaming applications, and the like.

Through the steps S202 to S208, a first object set including a plurality of objects having mutually exclusive relationships is obtained from the object to be pushed corresponding to the target account, the plurality of objects are screened according to the characteristics of the target account, one target object is screened out or none of the plurality of objects is selected, a second object set is obtained, the first object set in the object to be pushed is replaced by the second object set, so that the objects with mutual exclusive relationship in the objects to be pushed are only one of the target accounts or not pushed, thereby not only ensuring that a certain class of objects are recommended to proper users, but also distinguishing and distinguishing different user groups suitable for the objects with mutual exclusive relationship, therefore, the exposure utilization rate is improved, the commercial value of the display list is optimized, and effective pushing is carried out on mutually exclusive products.

Optionally, in this embodiment, the pushing method of the object to be pushed may be, but is not limited to, performed by a server for pushing the object. The server screens the objects to be pushed and pushes the screened results to the client, and the user can browse the contents pushed by the server on the client and perform operations such as purchasing and ordering.

It should be noted that, the objects having mutually exclusive relationship may be, but are not limited to, called as mutually exclusive objects, the number of mutually exclusive objects may be, but is not limited to, two or more, the following description uses the number of mutually exclusive objects as two for example, and similarly, the description is not repeated for the case where the number of mutually exclusive objects is more than two.

In the technical solution provided in step S202, the object to be pushed corresponding to the target account may be, but is not limited to, an object that is screened out for the target account according to account characteristics, history data, and other information of the target account and conforms to the preference of the target account.

Optionally, in this embodiment, the object to be pushed may include, but is not limited to: advertisements, insurance, audio-visual, articles, news, public numbers, applets, and so forth.

Optionally, in this embodiment, mutually having a mutual exclusion relationship means that only one of the plurality of objects is allowed to perform a mutual exclusion operation, where the mutual exclusion operation may include, but is not limited to: purchase operations, play operations, order placement operations, browse operations, and the like.

For example, any user only needs to purchase one object of the multiple objects, and the user may purchase two or more objects of the multiple objects at the same time without providing additional value for the user, which may cause waste and subsequent disputes. For example, one example in the area of insurance recommendations is the recommendation of medical insurance (medical insurance is a category of insurance items): one insurance product is a medical insurance oriented to high-end users and with higher premium and premium; the other is a medical insurance product with the same property of low premium and premium for users with low purchasing power. The users of the two insurance products should not repeatedly purchase the insurance products, because the two insurance products are both of the reimbursement type and cannot repeatedly settle the claims, namely, the two insurance products have mutually exclusive relationship.

Optionally, in this embodiment, the object to be pushed corresponding to the target account may be, but is not limited to, an object that is previously screened for the target account according to screening rules such as personalization and the same category and can be pushed to the target account. The personalized screening rule can be that different objects are screened for the account for each account according to the information such as the portrait characteristics and the historical behaviors of the account and the like to be displayed. The screening rules of the same category may mean that there are two similar objects to be pushed in the object list screened for the account, and the two have substantially the same function, so as to solve the same user requirement.

In the above step S202, the first set of objects may be obtained, but is not limited to, in the following manner:

s11, classifying the objects to be pushed according to a mutual exclusion relationship to obtain a plurality of object sets, wherein the mutual exclusion relationship represents that only one object in the plurality of objects is allowed to execute mutual exclusion operation;

s12, determining the object set whose number of objects included in the plurality of object sets is greater than 1 as the first object set.

Optionally, in this embodiment, classifying the objects to be pushed according to the mutual exclusion relationship may be to find out an object having the mutual exclusion relationship from the objects to be pushed, and divide the objects having the same mutual exclusion relationship into the same object set, for example: the objects to be pushed comprise an object 1, an object 2, … … and an object 8, wherein the object 1, the object 2 and the object 3 have mutual exclusion relationship, the object 4 and the object 6 have mutual exclusion relationship, the object 5 and the object 8 have mutual exclusion relationship, and the object 7 has no mutual exclusion relationship with other objects, so that the objects to be pushed can be divided into four object sets, namely an object set 11 (the object 1, the object 2 and the object 3), an object set 12 (the object 4 and the object 6), an object set 13 (the object 5 and the object 8) and an object set 14 (the object 7).

Optionally, in this embodiment, the object sets whose number of objects included in the plurality of object sets is greater than 1 may be, but not limited to, several object sets, and the several object sets are sequentially determined as the first object set to perform subsequent object screening operations. Such as: the object set 11 (object 1, object 2 and object 3), the object set 12 (object 4 and object 6) and the object set 13 (object 5 and object 8) are sequentially used as a first object set to screen the mutually exclusive relationship objects, and the obtained second object sets can be an object set 21 (object 2), an object set 22 (object 4), an object set 23 (object 8) and an object set 24 (empty set), respectively.

In the technical solution provided in step S204, the second object set obtained by filtering the plurality of objects in the first object set includes one target object matched with the target account characteristics in the plurality of objects, or the second object set is an empty set. That is, one object matching the target account feature may be selected from the plurality of objects and pushed as the target object, or one object may not be selected from the plurality of objects. For example: assuming that the two mutually exclusive objects included in the first set of objects are item a and item B, there are three possibilities for personalizing the list of items that are ultimately presented to the user on the item side: the user's item list contains only a, or the user's item list contains only B, or the user's item list contains neither a nor B.

In the step S204, the plurality of objects may be filtered by, but not limited to, the following method to obtain the second object set:

s21, calculating an operation probability corresponding to each object in the plurality of objects according to the target account characteristics, wherein the operation probability is used for indicating the probability that the target account performs a mutual exclusion operation on each object;

s22, screening the target objects with the operation probabilities meeting a target probability threshold from the plurality of objects, wherein the operation probabilities meeting the target probability threshold are indicated, and the probability of the target account performing the exclusive operation on the target objects is higher than the probability of performing the exclusive operation on the objects except the target object in the plurality of objects;

s23, determining that the second object set comprises the target object under the condition that the target object is screened out;

and S24, determining the second object set to be an empty set under the condition that the target object is not screened out.

Optionally, in this embodiment, the operation probability is used to indicate a probability that the target account performs a mutual exclusion operation on each object, and the operation probability may be predicted according to the characteristics of the target account.

Optionally, in this embodiment, the mutex operation may refer to, but is not limited to, an operation that a user only needs to perform on one of the mutex objects, such as: for objects such as audio and video, advertisements and the like, the mutual exclusion operation can be but is not limited to playing operation, and for objects such as articles, products and the like, the mutual exclusion operation can be but is not limited to purchasing operation or ordering operation and the like.

Optionally, in this embodiment, if a target object whose operation probability satisfies the target probability threshold is screened from the multiple objects, the target object is included in the second object set, and if a target object whose operation probability satisfies the target probability threshold is not screened from the multiple objects, the second object set is an empty set.

Optionally, in this embodiment, but not limited to, a corresponding target probability threshold may be determined for each object in the plurality of objects, if the operation probability of a certain object satisfies the corresponding target probability threshold, the certain object is determined as a target object, and if the operation probabilities of all the objects do not satisfy the corresponding target probability threshold, the second object set is an empty set. Such as: for two mutually exclusive object items A and B in a shopping scene, if the operation probability satisfies the condition A, the item A is recommended for the user, that is, the second object set comprises the item A. If the operation probability satisfies the condition B, the item B is recommended for the user, that is, the second object set comprises the item B. If the operation probability neither satisfies the condition a nor the condition B, neither the item a nor the item B is recommended, that is, the second set of objects is an empty set.

In the above step S21, the operation probability may be calculated by, but not limited to:

s31, respectively inputting the target account characteristics into a first model corresponding to each of the plurality of objects and a second model corresponding to a plurality of objects having a mutual exclusion relationship, where the first model is configured to output, according to the input account characteristics, a probability that an account performs the mutual exclusion operation on each of the plurality of objects in a scenario where the each of the plurality of objects is displayed in an account, and the second model is configured to respectively output, according to the input account characteristics, a plurality of probabilities that an account performs the mutual exclusion operation on the plurality of objects in a scenario where the plurality of objects are simultaneously displayed in an account;

s32, obtaining a first probability of each first model output to obtain a plurality of first probabilities, and obtaining a plurality of second probabilities of the second model output;

s33, determining the operation probability according to the plurality of first probabilities and the plurality of second probabilities.

Optionally, in this embodiment, the first model is configured to output, according to the input account characteristic, a probability that the account performs a mutual exclusion operation on each object in a scene where each object is shown in the account. Taking two mutually exclusive type objects a and B as an example, the first model may include two sub-models, model a and model B, which correspond to the two mutually exclusive type objects, respectively, and input the target account characteristics into model a, model a outputs the probability of performing the mutual exclusion operation on a, and inputs the target account characteristics into model B, and model B outputs the probability of performing the mutual exclusion operation on B.

Alternatively, in this embodiment, in a shopping scenario, the first models may be referred to as single item purchase intention models, the output of each first model represents the probability that the user purchases the corresponding item, and the higher the output value, the more likely the user purchases the corresponding item.

It should be noted that, the description taking the shopping scenario as an example in this embodiment is only an example in this embodiment, and does not limit the application scenario of this embodiment, and the applications in other scenarios are similar to this, and are not described herein again.

Optionally, in this embodiment, the second model is configured to output, according to the input account characteristics, a plurality of probabilities that the account performs a mutual exclusion operation on the plurality of objects in a scenario where the plurality of objects are simultaneously displayed in the account. Taking two mutually exclusive objects A and B as an example, the second model can be a model AB, the target account characteristics are input into the model AB, and the model AB can respectively output the probability of executing the mutually exclusive operation on A and the probability of executing the mutually exclusive operation on B.

Optionally, in this embodiment, in a shopping scenario, the second model may be referred to as a mutually exclusive item purchase intention model, where multiple probability values output by the second model represent relative probabilities of a user purchasing one of multiple items when the user simultaneously displays the items, and a higher probability value indicates that the user is more likely to purchase the item when selecting one more item.

Optionally, in this embodiment, the plurality of probability values output by the second model are a set of relative probability values, which sum to 1, each probability value representing a probability of several percent purchasing an item in case of multiple items being presented simultaneously.

Optionally, in this embodiment, the process of determining the operation probability according to the first probabilities and the second probabilities may be to calculate a probability value as the operation probability, or calculate a plurality of probability values to represent the operation probability. The operation probability satisfying the target probability threshold value indicates that the probability that the target account performs the exclusive operation on the target object is higher than the probability that the objects except the target object in the plurality of objects perform the exclusive operation. Therefore, the operation probability may represent a probability that the target account performs the mutual exclusion operation on each object, and may be a probability value that is higher or lower than the probability that the target account performs the mutual exclusion operation on each object.

Such as: taking two mutually exclusive objects a and B as an example, the operation probability may include: the probability of performing a mutual exclusion operation on A, the probability of performing a mutual exclusion operation on B, and the probability of performing a mutual exclusion operation on A and on B. The probability of executing the mutual exclusion operation on the A represents the probability of executing the mutual exclusion operation on the A by the target account when the A is displayed by the target account, the probability of executing the mutual exclusion operation on the B represents the probability of executing the mutual exclusion operation on the B by the target account when the B is displayed by the target account, and the probability of executing the mutual exclusion operation on the A and the B represents two probabilities of executing the mutual exclusion operation on the A and the B by the target account when the A and the B are displayed by the target account.

As an alternative embodiment, before the step S31, the method further includes:

s41, training a first initial model by using a first feature sample corresponding to each object to obtain a first model corresponding to each object, where the first feature sample is a feature sample of an account for executing the mutual exclusion operation on each object in an operation scene corresponding to each object;

s42, training a second initial model by using a second feature sample corresponding to the multiple objects to obtain the second model, where the second feature sample is a feature sample of an account performing the exclusive operation on one object of the multiple objects in an operation scenario corresponding to the multiple objects.

Optionally, in this embodiment, each model may be trained, but is not limited to, using the following training process: the method comprises the steps that all characteristic samples used in training are provided with labels, a loss function is configured for each model in advance, the characteristic samples are input into an initial model to obtain output of the initial model, the output of the initial model and the labels carried by the characteristic samples are substituted into the loss function to be calculated, model parameters of the initial model are adjusted according to calculation results, the characteristic samples are input into the model after parameter adjustment to obtain output of the model, and the circulation is carried out until the output of the model and the labels carried by the characteristic samples are substituted into the loss function to be calculated, and the obtained model is used as a target model when the result obtained by calculation meets preset end conditions.

Optionally, in this embodiment, for different models, the sample data may originate from different operation scenarios, but the corresponding scenarios need to be matched with the corresponding models to be trained. For a model corresponding to a single object (i.e. the first model described above), the corresponding first feature sample may be any exposure list containing historical exposure data of the object, such as details page access user data of the object, and so on. And for the one-out-of-two model or the one-out-of-more model (i.e., the second model) of the object, the second feature sample may select the scene data equally exposed by each object. The sample of the one-out-of-two or more-out-of-one model may be taken from a non-recommended scene, for example, a second feature sample may be taken from a user who has presented both objects (not necessarily the same scene) on the same day.

In an alternative embodiment, taking an item transaction scenario as an example, the offline model training may first classify the user into a customer group based on historical exposure data (e.g., purchase data of the user), and taking item a and item B as mutually exclusive objects, according to the purchase condition of the user for two mutually exclusive items, classify the user into 4 categories: a customer category A of users who purchased only item A; a customer category B of users who purchased only item B; a customer category C of users who purchased both item A and item B; customer category N for users who have not purchased both items. For strict mutually exclusive articles, if the scheme is adopted to limit the service strategy, the class C of the guest group may not exist. Based on the above crowd classification, 3 different types of models can be constructed, as shown in table 1, the model a and the model B are single item purchase intention models (i.e., the above first model), the output of the model represents the probability that the user purchases the corresponding item, and the higher the output of the model is, the more likely the user purchases the corresponding item; the model AB is a tendency model (i.e. the second model) between mutually exclusive articles, the output of the model represents the relative probability of the user purchasing one of the two articles when the user simultaneously displays the two articles, and the score is higher when the relative probability of the model AB corresponding to the purchased article a is set, so that the user is more likely to purchase the article a when the user selects one of the articles.

Optionally, in this embodiment, for the alternative model, the probabilities corresponding to the two items may be output. Or only the probability corresponding to one article can be output, and then the probability corresponding to another article can be calculated through the output of the 1-model. And for the one-out-of-multiple model, taking the multiple probabilities as the output of the model, and providing the corresponding relation between the output probability and the corresponding article.

TABLE 1

The model can be constructed based on user purchase data, and if the purchase is replaced by clicking on each article by the user, a set of models with similar effects can be obtained, such as: the feature samples used for training the model may include user purchase data, user click data, or more complex object construction strategies may be applied, such as combining click and purchase construction features at the same time, and then the feature samples used for training the model may include user purchase data and user click data at the same time, or different weights may be applied to different samples for different scenarios and parameters (such as premium size and price), such as: in the scenario of policy recommendation, different feature samples have weights proportional to the premium size according to the corresponding premium size, so that the screening effect can be further improved.

In the above step S33, the operation probability may be determined by, but is not limited to, the following manner:

s51, for each probability in the first probabilities and the second probabilities, determining the ranking position of each probability in a target probability set, wherein the target probability set records the historical probability of the historical account performing the mutual exclusion operation on the object corresponding to each probability under the scene corresponding to each probability;

s52, determining the operation probability according to the sorting position.

Optionally, in this embodiment, the operation probability corresponding to each probability is re-determined by using the statistical ranking condition of each current probability in the historical probabilities, and the dynamic probability ranking of each user can be calculated according to the probabilities output by each original model of each user and the statistical value of the overall probability of the historical account, and the operation probability of the user is calculated according to the ranking. Therefore, the absolute numerical value output by the model is converted into a relative numerical value compared with the online user, and the influence of deviation caused by the absolute value on the recommendation result is avoided.

Optionally, in this embodiment, the ranking position of each probability in the target probability set may be determined by, but is not limited to, the following method:

s61, inputting each probability into a database for storing the target probability set;

s62, sorting the historical probabilities and each probability stored in the target probability set through the database;

s63, dividing the ordered target probability set into sub probability sets with target quantity, wherein each sub probability set corresponds to a set serial number;

s64, determining the sub-probability set in which each probability falls as a target sub-probability set, and obtaining the sequencing position.

Optionally, in this embodiment, the operation probability may be determined according to the ranking position in the following manner, but is not limited to:

and S65, determining the target set sequence number corresponding to the target sub probability set as the operation probability corresponding to each probability.

Optionally, in this embodiment, the database for storing the target probability set may be, but is not limited to, a database with an automatic ordering property, such as: and the Redis database automatically sequences the originally stored historical probabilities and each newly input probability in the database by utilizing the automatic sequencing characteristic of the Redis database.

Optionally, in this embodiment, the manner of dividing the sub-probability sets may be uniform division, or division may be performed according to the distribution of the probabilities stored in the target probability set.

Optionally, in this embodiment, the position of each sub probability set arranged in all sub probability sets represents the relative position of the probability included in the sub probability sets in the overall probability. The position may be represented by a collection number, which may be a preset number, such as: from 1 to 100, from 0.01 to 1, may also be determined from the probabilities stored in the set of sub-probabilities, such as: it may be the maximum probability, the minimum probability or the mean of the probabilities stored in the sub-probability set, etc.

Optionally, in this embodiment, the sorted target probability sets may be divided into the target number of sub-probability sets by, but not limited to, the following ways:

s71, averagely dividing the probability included in the sorted target probability set into the sub probability sets of the target number;

s72, mapping the sub-probability sets of the target number to a preset value interval, wherein the value interval comprises the value of the target number, and the value of the target number corresponds to the sub-probability sets of the target number one by one;

and S73, determining the numerical value corresponding to each sub probability set as the set serial number corresponding to each sub probability set.

Optionally, in this embodiment, the sub-probability set partitioning process may be, but is not limited to, a mapping between numerical ranges, such as: a probability value in the range of 0 to 1 is mapped to a value interval of 1 to 100. The probabilities included in the ordered target probability set are equally divided into 100 sub-probability sets, and the sub-probability set arranged in the order of the few bits is mapped to the number of the preset value interval 1 to 100, such as: the sub-probability set ranked at bit 49 is mapped to the value 49 in the preset value range 1 to 100. Thereby obtaining the set sequence number corresponding to each sub-probability set.

Optionally, in this embodiment, the process of using the set sequence number as the operation probability by the partition set may be, but is not limited to, referred to as an adaptive bucket partitioning process. The operation probability used in the object screening process can be constructed not based on the original output of each model directly, but an adaptive bucket dividing mechanism is introduced, the operation probability is constructed based on the ranking quantiles output by the adaptive bucket dividing mechanism, and the advantages of introducing the adaptive bucket dividing mechanism can be embodied in the following three aspects:

firstly, because the original models are all two-classification models, the output is the relative probability of the two-classification models, the size of the two-classification models is directly related to the proportion of positive and negative samples of model samples, the two-classification models can carry out specific selection and optimization to ensure the model effect, the proportion of the positive and negative samples and the weights of different samples, the model scoring can be directly influenced by the proportion of the positive and negative samples and the weight of the samples, the absolute value of the model scoring is unstable, a self-adaptive bucket-dividing mechanism is introduced, the relative size relation between the scoring is extracted, and the influence of the absolute size of the scoring can be effectively removed.

Secondly, the preference of mutually exclusive article that the nature is approximate just can be distinguished to the alternative model need have very high resolving power, so the model scoring absolute value receives the very easy influence that is unstable of sample characteristic, and the original scoring of direct application, this kind of unstability can directly conduct on-line, causes the unstability of online recommendation result, introduces the self-adaptation and divides the bucket mechanism, can make online recommendation result stability stronger.

Third, vertical domain recommendations need to take into account variability in business goals. Sometimes, based on the consideration of the overall business value, it is necessary to increase the exposure rate of a certain article, for example, to increase the expected exposure rate of a certain article from 20% to 30%, to achieve this effect, at the system level, the threshold of the online policy needs to be adjusted, and if the original score is directly applied, the threshold also corresponds to the original score, and does not directly correspond to the index of the business target layer, and the threshold is made by the value of the original score type, which may cause the instability of the business target index.

Based on the above three points, in this embodiment, an adaptive bucket partitioning mechanism is introduced between the offline model output probability and the online policy (comparison between the operation probability and the target probability threshold), and its function is to calculate the dynamic probability ranking of each user according to the statistical value of each user's original model probability and the historical account overall probability, and calculate the ranking quantiles of the users according to the ranking as the comparison between their corresponding operation probability and the target probability threshold. The adaptive bucket dividing mechanism divides (for example, can evenly divide) the historical user and the current user into a certain number of (for example, 100) buckets according to the ranking, and each user outputs the bucket dividing sequence number of each model as the operation probability.

Fig. 3 is a schematic diagram of an adaptive bucket partitioning mechanism according to an embodiment of the present application, and as shown in fig. 3, an object push system online service first requests to obtain a plurality of first probabilities and a plurality of second probabilities output by a plurality of first models and a second model, then writes each probability value into a Redis database (memory phase), then obtains a sorting sequence number of a current probability in a user pool by using an automatic sorting characteristic of the Redis database, calculates a bucket score (bucket partitioning phase) according to the number of users in the user pool and the sorting sequence number, and finally outputs a sequence number (rank) value of a bucket as a final operation probability.

Optionally, in this embodiment, before determining the ranking position of each probability in the target probability set, the data stored in the database may be filtered in the following manner, but is not limited to:

s81, acquiring an initial probability set corresponding to each probability;

s82, deleting the probability that the storage duration exceeds the duration threshold from the initial probability set to obtain the target probability set.

Optionally, in this embodiment, another reason for selecting the Redis database to implement automatic bucket allocation is that the above-mentioned obsolete process of expiration can be implemented on data by using an expiration mechanism of the Redis database, and dynamic update of the user pool of the adaptive bucket allocation is maintained.

As an alternative embodiment, the screening the target object whose operation probability satisfies the target probability threshold from the plurality of objects includes one of:

s91, determining the object with the operation probability corresponding to the first probability higher than a first threshold and the operation probability corresponding to the second probability higher than a second threshold as the target object;

s92, determining the object as the target object, wherein the operation probability corresponding to the first probability is higher than a third threshold, the operation probabilities corresponding to the first probabilities of the other objects are lower than a fourth threshold, and the operation probability corresponding to the second probability is higher than a fifth threshold.

Optionally, in this embodiment, whether the operation probability satisfies the target probability threshold may be, but is not limited to, referred to as an online policy, which is used to determine whether and which object to push. The goal of online strategic optimization is single user value, rather than simple purchase conversion rate, because the user value of a single order for different items is different: for example, item B, which is a low end user oriented product, is much less valuable than item a, which is a high end user oriented product, and, for the same exposure amount, item B requires more user interaction to earn out on item a or beyond a.

An online policy may be to screen out an object whose operation probability corresponding to the first probability is higher than a first threshold (indicating that the user has a stronger intention to purchase the object) and whose operation probability corresponding to the second probability is higher than a second threshold (indicating that the user has a stronger intention to purchase the object than to purchase other objects), where the online policy may indicate that the user has a stronger intention to purchase the object in a single scene in which the object is recommended, and that the user has a stronger intention to purchase the object in a scene in which a plurality of mutually exclusive objects are recommended. An alternative expression of the above-described on-line strategy is given, as shown in table 2.

TABLE 2

The conditions given in table 2 are a simpler on-line strategy applying multiple model bucket scoring, where the conditions for each different product form are of definite physical significance, for example, the conditions for PA can be interpreted as: only users who have a willingness to purchase item a above a specified threshold, threshold1, while being more willing to purchase a among items a and B, i.e., a relative willingness to purchase greater than threshold2, push item a; and PB is similarly conditioned.

Optionally, in this embodiment, before screening the target object whose operation probability satisfies the target probability threshold from the plurality of objects, the target probability threshold may be determined in, but is not limited to, the following manner:

s101, exhausting all candidate probability threshold combinations allowed to appear under the condition of screening the plurality of objects;

s102, calculating an evaluation parameter corresponding to each candidate probability threshold combination under the condition that each candidate probability threshold combination in all the candidate probability threshold combinations is taken as a screening condition for screening the plurality of objects, wherein the evaluation parameter is used for indicating the value of the object generated under the screening condition of each candidate probability threshold combination;

s103, determining the candidate probability threshold combination with the highest object value in all the candidate probability threshold combinations as the target probability threshold.

Optionally, in this embodiment, the evaluation parameter is used to indicate the value of the object generated under the filtering condition of each candidate probability threshold combination, such as: the evaluation parameter may be used to indicate a composite value (which may be, but is not limited to, indicated by the evaluation parameter) under a certain screening condition. The evaluation parameters may include, but are not limited to, resource growth values that may be used to represent, but are not limited to, the difference between the revenue generated by the operating process of the subject and the cost of the subject, such as: in a trading scenario, the resource growth value may be represented, but is not limited to, using an average profit, a total profit, a net profit, a gross profit, and the like.

Optionally, in this embodiment, the calculating the evaluation parameter corresponding to each candidate probability threshold combination may include, but is not limited to, the following:

s111, taking each candidate probability threshold combination as a screening condition, and screening the plurality of objects for the account sample to obtain a screening result;

s112, calculating the recall rate of each object in the plurality of objects according to the screening result;

s113, estimating the conversion rates of the plurality of objects according to the recall rate;

s114, calculating the resource growth value generated by the account sample according to the conversion rate of the plurality of objects as the evaluation parameter.

Such as: 4 pending parameters in the above online policy: the determination of threshold1, threshold2, threshold3, and threshold4 can be combined to take into account the on-line effect and specific goals, and a simple and straightforward way to determine the threshold is to calculate the recall rate of each product by exhaustively enumerating different combinations of thresholds, and based on this, combine historical conversion rates to calculate a combined goal (such as average profit), and select the optimal combination of thresholds. Because the threshold value corresponds to the score of the barrel-dividing score, the solution space of the combination of the three models is not more than 100 ten thousand, and the exhaustive search of the solution space with the magnitude order is completely feasible, which is one of the benefits brought by the online introduction of the self-adaptive barrel-dividing module.

The online policy is not limited to the simple form of table 2, the filtering rule of each product form may integrate a combination of a plurality of scores and a plurality of conditions, and another online policy may be to screen out an object whose first probability corresponds to an operation probability higher than a third threshold, and whose first probability corresponds to an operation probability lower than a fourth threshold and whose second probability corresponds to an operation probability higher than a fifth threshold. For example, for the product form PA, the following conditions are also a reasonable hit rule:

MA_rank>threshold5，MB_rank<threshold6 and MAB_rank>threshold7。

the online policy of the PA applied in practice may be a combination of the rules as described above, and the combination of a plurality of rules generally improves the recall and overall effect of the system, and is similar to the establishment of online policies for other product forms. When the rule is complex, it is not suitable to apply an exhaustive strategy to determine the threshold parameter in the rule, and at this time, the optimal threshold combination can be learned through a machine learning algorithm. For example, the condition for PA can be interpreted as: the performance of the article A is more suitable for the characteristics of the user, and the purchase intention of the user for the article A is stronger, the object is used for training a machine learning algorithm to obtain a machine learning model so as to filter whether the article A is recommended or not for the user.

In the technical solution provided in step S206, the first object set is replaced by the second object set in the object to be pushed, so that the objects having a mutual exclusion relationship in the object to be pushed only include one or none of the objects in the target pushed object.

Such as: the objects to be pushed comprise objects 1, 2, … … and 8, wherein the first object set comprises an object set 11 (objects 1, 2 and 3), an object set 12 (objects 4 and 6), an object set 13 (objects 5 and 8) and an object set 14 (objects 7), and the second object set comprises an object set 21 (objects 2), an object set 22 (objects 4), an object set 23 (objects 8) and an object set 24 (empty set) after screening. Then after the replacement, the target push objects include object 2, object 8, and object 4.

In the technical solution provided in step S208, the manner of pushing the target push object to the target account may include, but is not limited to: and sending the target push object to an instant messaging session of the target account by using an instant messaging application, sending the target push object to a social interaction platform of the target account, sending the target push object to a private letter of the target account, and the like.

The application further provides an optional embodiment, the optional embodiment provides a product recommendation system, fig. 4 is a schematic diagram of a system architecture of the product recommendation system according to the optional embodiment of the application, as shown in fig. 4, the system offline trains a plurality of models with different properties based on click and purchase behavior data of online users, for each online user, each model outputs an original model score based on user characteristics, each model score is transformed by an online adaptive bucket-dividing module to obtain a corresponding ranking quantile, and then an online policy module determines whether a specific condition is met based on combination of the ranking quantiles to determine a final article display policy.

It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present application is not limited by the order of acts described, as some steps may occur in other orders or concurrently depending on the application. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required in this application.

Through the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but the former is a better implementation mode in many cases. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling an electronic device (such as a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present application.

According to another aspect of the embodiment of the present application, there is also provided a pushing apparatus for an object to be pushed, which is used for implementing the pushing method for an object to be pushed. Fig. 5 is a schematic diagram of an alternative pushing apparatus for an object to be pushed according to an embodiment of the present application, and as shown in fig. 5, the apparatus may include:

an obtaining module 52, configured to obtain a first object set from objects to be pushed corresponding to a target account, where the first object set includes a plurality of objects having mutually exclusive relationships;

a screening module 54, configured to screen the multiple objects according to target account characteristics of the target account to obtain a second object set, where the second object set includes one of the multiple objects that matches the target account characteristics, or the second object set is an empty set;

a replacing module 56, configured to replace the first object set in the object to be pushed with the second object set, so as to obtain a target pushed object;

a pushing module 58, configured to push the target pushing object to the target account.

It should be noted that the obtaining module 52 in this embodiment may be configured to execute the step S202 in this embodiment, the screening module 54 in this embodiment may be configured to execute the step S204 in this embodiment, the replacing module 56 in this embodiment may be configured to execute the step S206 in this embodiment, and the pushing module 58 in this embodiment may be configured to execute the step S208 in this embodiment.

It should be noted here that the modules described above are the same as the examples and application scenarios implemented by the corresponding steps, but are not limited to the disclosure of the above embodiments. It should be noted that the modules described above as a part of the apparatus may operate in a hardware environment as shown in fig. 1, and may be implemented by software or hardware.

Through the module, a first object set comprising a plurality of objects with mutual exclusive relation is obtained from the objects to be pushed corresponding to the target account, the objects are screened according to the characteristics of the target account, one target object is screened out or none of the objects is selected, a second object set is obtained, the second object set is used for replacing the first object set in the objects to be pushed, so that the objects with mutual exclusive relationship in the objects to be pushed are only one of the target accounts or not pushed, thereby not only ensuring that a certain class of objects are recommended to proper users, but also distinguishing and distinguishing different user groups suitable for the objects with mutual exclusive relationship, therefore, the exposure utilization rate is improved, the commercial value of the display list is optimized, and effective pushing is carried out on mutually exclusive products.

Optionally, the screening module comprises:

a calculating unit, configured to calculate, according to the target account characteristics, an operation probability corresponding to each object in the multiple objects, where the operation probability is used to indicate a probability that the target account performs a mutual exclusion operation on each object;

a screening unit, configured to screen the target object with the operation probability satisfying a target probability threshold from the plurality of objects, where the operation probability satisfying the target probability threshold is indicated, and the probability that the target account performs the mutually exclusive operation on the target object is higher than the probability that an object other than the target object performs the mutually exclusive operation on the plurality of objects;

a first determining unit, configured to determine that the second object set includes the target object when the target object is screened out;

and the second determining unit is used for determining that the second object set is an empty set under the condition that the target object is not screened out.

Optionally, the computing unit is configured to:

respectively inputting the target account characteristics into a first model corresponding to each object in the plurality of objects and a second model corresponding to the plurality of objects with a mutual exclusion relationship, wherein the first model is used for outputting the probability of executing the mutual exclusion operation on each object by an account under the scene of displaying each object to the account according to the input account characteristics, and the second model is used for respectively outputting a plurality of probabilities of executing the mutual exclusion operation on the plurality of objects by the account under the scene of simultaneously displaying the plurality of objects to the account according to the input account characteristics;

obtaining a first probability of each first model output to obtain a plurality of first probabilities, and obtaining a plurality of second probabilities of the second model output;

determining the operational probability based on the plurality of first probabilities and the plurality of second probabilities.

Optionally, the computing unit is configured to:

for each probability in the plurality of first probabilities and the plurality of second probabilities, determining a ranking position of the each probability in a target probability set, wherein the target probability set records a historical probability that a historical account performs the mutual exclusion operation on an object corresponding to the each probability under a scene corresponding to the each probability;

and determining the operation probability according to the sequencing position.

Optionally, the computing unit is configured to: inputting said each probability into a database for storing said target set of probabilities; ranking, by the database, the historical probabilities and each of the probabilities stored in the target probability set; dividing the ordered target probability set into sub probability sets with target quantity, wherein each sub probability set corresponds to a set serial number; determining the sub-probability set in which each probability falls as a target sub-probability set to obtain the sequencing position; and determining the target set sequence number corresponding to the target sub-probability set as the operation probability corresponding to each probability.

Optionally, the computing unit is configured to:

averagely dividing the probability included in the ordered target probability set into the sub probability sets of the target number;

mapping the sub-probability sets of the target number to a preset value interval, wherein the value interval comprises the values of the target number, and the values of the target number correspond to the sub-probability sets of the target number one by one;

and determining the numerical value corresponding to each sub-probability set as the set serial number corresponding to each sub-probability set.

Optionally, the apparatus further comprises:

an obtaining module, configured to obtain an initial probability set corresponding to each probability before determining a ranking position of each probability in a target probability set;

and the deleting module is used for deleting the probability that the storage duration exceeds a duration threshold from the initial probability set to obtain the target probability set.

Optionally, the apparatus further comprises:

a first training module, configured to train a first initial model using a first feature sample corresponding to each object to obtain a first model corresponding to each object before inputting the target account features into a first model corresponding to each object and a second model corresponding to the plurality of objects, where the first feature sample is a feature sample of an account for executing the mutually-exclusive operation on each object in an operation scene corresponding to each object;

a second training module, configured to train a second initial model using a second feature sample corresponding to the multiple objects to obtain a second model before inputting the target account features into the first model corresponding to each of the multiple objects and the second model corresponding to the multiple objects, where the second feature sample is a feature sample of an account for performing the mutual exclusion operation on one of the multiple objects in an operation scene corresponding to the multiple objects.

Optionally, the screening unit is for one of:

determining the object with the operation probability corresponding to the first probability higher than a first threshold and the operation probability corresponding to the second probability higher than a second threshold as the target object;

and determining the object with the operation probability corresponding to the first probability higher than a third threshold value, the operation probabilities corresponding to the first probabilities of other objects lower than a fourth threshold value and the operation probabilities corresponding to the second probabilities higher than a fifth threshold value as the target object.

Optionally, the apparatus further comprises:

an exhaustion module configured to exhaust all candidate probability threshold combinations allowed to occur in the case of screening the plurality of objects before screening the target objects of which the operation probabilities satisfy a target probability threshold from the plurality of objects;

a calculating module, configured to calculate an evaluation parameter corresponding to each candidate probability threshold combination when each candidate probability threshold combination in all the candidate probability threshold combinations is used as a screening condition for screening the multiple objects, where the evaluation parameter is used to indicate a value of an object generated under the screening condition of each candidate probability threshold combination;

and the determining module is used for determining the candidate probability threshold combination with the highest object value in all the candidate probability threshold combinations as the target probability threshold.

Optionally, the computing module is configured to:

taking each candidate probability threshold combination as a screening condition, and screening the plurality of objects for the account sample to obtain a screening result;

calculating the recall rate of each object in the plurality of objects according to the screening result;

estimating conversion rates of the plurality of objects according to the recall rate;

and calculating the resource growth value generated by the account sample according to the conversion rate of the plurality of objects as the evaluation parameter.

Optionally, the obtaining module includes:

the classification unit is used for classifying the objects to be pushed according to a mutual exclusion relationship to obtain a plurality of object sets, wherein the mutual exclusion relationship represents that the mutual exclusion operation is allowed to be executed on only one object in the plurality of objects;

a third determining unit, configured to determine, as the first object set, an object set in which the number of objects included in the plurality of object sets is greater than 1.

It should be noted here that the modules described above are the same as the examples and application scenarios implemented by the corresponding steps, but are not limited to the disclosure of the above embodiments. It should be noted that the modules described above as a part of the apparatus may be operated in a hardware environment as shown in fig. 1, and may be implemented by software, or may be implemented by hardware, where the hardware environment includes a network environment.

According to another aspect of the embodiment of the present application, there is also provided an electronic device for implementing the pushing method for the object to be pushed.

Fig. 6 is a block diagram of an electronic device according to an embodiment of the present application, and as shown in fig. 6, the electronic device may include: one or more processors 601 (only one of which is shown), a memory 603, and a transmission device 605, as shown in fig. 6, the electronic device may further include an input-output device 607.

The memory 603 may be configured to store a software program and a module, such as a program instruction/module corresponding to the method and apparatus for pushing an object to be pushed in the embodiment of the present application, and the processor 601 executes various functional applications and data processing by running the software program and the module stored in the memory 603, that is, implements the method for pushing an object to be pushed. The memory 603 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 603 may further include memory located remotely from the processor 601, which may be connected to the electronic device through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The above-mentioned transmission device 605 is used for receiving or sending data via a network, and may also be used for data transmission between a processor and a memory. Examples of the network may include a wired network and a wireless network. In one example, the transmission device 605 includes a Network adapter (NIC) that can be connected to a router via a Network cable and other Network devices to communicate with the internet or a local area Network. In one example, the transmission device 605 is a Radio Frequency (RF) module, which is used for communicating with the internet in a wireless manner.

Among them, the memory 603 is used to store an application program, in particular.

The processor 601 may call the application stored in the memory 603 through the transmission device 605 to perform the following steps:

and pushing the target pushing object to the target account.

By adopting the embodiment of the application, a scheme for pushing the object to be pushed is provided. The method comprises the steps of acquiring a first object set comprising a plurality of objects with mutual exclusive relations from objects to be pushed corresponding to a target account, screening the objects according to the characteristics of the target account, screening out one target object or not selecting the objects to obtain a second object set, replacing the first object set in the objects to be pushed with the second object set, and enabling the objects with mutual exclusive relations in the objects to be pushed to be only pushed to one of the target accounts or not to be pushed to the target account.

Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment is not described herein again.

It will be understood by those skilled in the art that the structure shown in fig. 6 is merely an illustration, and the electronic device may be a smart phone (e.g., an Android phone, an iOS phone, etc.), a tablet computer, a palm computer, a Mobile Internet Device (MID), a PAD, etc. Fig. 6 is a diagram illustrating a structure of the electronic device. For example, the electronic device may also include more or fewer components (e.g., network interfaces, display devices, etc.) than shown in FIG. 6, or have a different configuration than shown in FIG. 6.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by a program for instructing hardware associated with an electronic device, where the program may be stored in a computer-readable storage medium, and the storage medium may include: flash disks, Read-Only memories (ROMs), Random Access Memories (RAMs), magnetic or optical disks, and the like.

Embodiments of the present application also provide a storage medium. Alternatively, in this embodiment, the storage medium may be a program code for executing a push method of an object to be pushed.

Optionally, in this embodiment, the storage medium may be located on at least one of a plurality of network devices in a network shown in the above embodiment.

Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps:

and pushing the target pushing object to the target account.

Optionally, in this embodiment, the storage medium may include, but is not limited to: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.

The above-mentioned serial numbers of the embodiments of the present application are merely for description and do not represent the merits of the embodiments.

The integrated unit in the above embodiments, if implemented in the form of a software functional unit and sold or used as a separate product, may be stored in the above computer-readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or a part of or all or part of the technical solution contributing to the prior art may be embodied in the form of a software product stored in a storage medium, and including instructions for causing one or more computer devices (which may be personal computers, servers, network devices, or the like) to execute all or part of the steps of the method described in the embodiments of the present application.

In the above embodiments of the present application, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the several embodiments provided in the present application, it should be understood that the disclosed client may be implemented in other manners. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one type of division of logical functions, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The foregoing is only a preferred embodiment of the present application and it should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present application, and these improvements and modifications should also be considered as the protection scope of the present application.

Claims

1. A method for pushing an object to be pushed is characterized by comprising the following steps:

and pushing the target pushing object to the target account.

2. The method of claim 1, wherein the screening the plurality of objects according to the target account characteristics of the target account to obtain a second set of objects comprises:

calculating an operation probability corresponding to each object in the plurality of objects according to the characteristics of the target account, wherein the operation probability is used for indicating the probability that the target account executes mutually exclusive operation on each object;

screening the target objects with the operation probabilities meeting a target probability threshold value from the plurality of objects, wherein the operation probabilities meeting the target probability threshold value, and the probability of the target account performing the mutual exclusion operation on the target objects is higher than the probability of performing the mutual exclusion operation on the objects except the target objects in the plurality of objects;

determining that the second object set comprises the target object under the condition that the target object is screened out;

determining that the second set of objects is an empty set if the target object is not screened out.

3. The method of claim 2, wherein calculating the probability of operation for each of the plurality of objects based on the target account characteristics comprises:

4. The method of claim 3, wherein determining the operational probability based on the plurality of first probabilities and the plurality of second probabilities comprises:

and determining the operation probability according to the sequencing position.

5. The method of claim 4,

determining the ranked position of each probability in the target probability set comprises: inputting said each probability into a database for storing said target set of probabilities; ranking, by the database, the historical probabilities and each of the probabilities stored in the target probability set; dividing the ordered target probability set into sub probability sets with target quantity, wherein each sub probability set corresponds to a set serial number; determining the sub-probability set in which each probability falls as a target sub-probability set to obtain the sequencing position;

determining the operational probability according to the rank position comprises: and determining the target set sequence number corresponding to the target sub-probability set as the operation probability corresponding to each probability.

6. The method of claim 5, wherein dividing the sorted target probability sets into a target number of sub-probability sets comprises:

7. The method of claim 4, wherein prior to determining the ranked position of each probability in the target probability set, the method further comprises:

acquiring an initial probability set corresponding to each probability;

and deleting the probability that the storage duration exceeds a duration threshold from the initial probability set to obtain the target probability set.

8. The method of claim 3, wherein prior to entering the target account features into the first model for each of the plurality of objects and the second model for the plurality of objects, the method further comprises:

training a first initial model by using a first feature sample corresponding to each object to obtain a first model corresponding to each object, wherein the first feature sample is a feature sample of an account for executing the mutual exclusion operation on each object in an operation scene corresponding to each object;

and training a second initial model by using a second feature sample which corresponds to the plurality of objects together to obtain the second model, wherein the second feature sample is a feature sample of an account which executes the mutual exclusion operation on one object in the plurality of objects under an operation scene which corresponds to the plurality of objects together.

9. The method of claim 3, wherein filtering the target object from the plurality of objects for which the operation probability satisfies the target probability threshold comprises one of:

10. The method of claim 2, wherein prior to filtering the target objects from the plurality of objects for which the operational probability satisfies a target probability threshold, the method further comprises:

enumerating all candidate probability threshold combinations that are allowed to occur if the plurality of objects are screened;

calculating an evaluation parameter corresponding to each candidate probability threshold combination under the condition that each candidate probability threshold combination in all the candidate probability threshold combinations is used as a screening condition for screening the plurality of objects, wherein the evaluation parameter is used for indicating the value of the object generated under the screening condition of each candidate probability threshold combination;

and determining the candidate probability threshold combination with the highest object value in all the candidate probability threshold combinations as the target probability threshold.

11. The method of claim 10, wherein calculating the evaluation parameter corresponding to each candidate probability threshold combination comprises:

12. The method of claim 1, wherein obtaining the first set of objects from the objects to be pushed corresponding to the target account comprises:

classifying the objects to be pushed according to a mutual exclusion relationship to obtain a plurality of object sets, wherein the mutual exclusion relationship represents that only one object in the plurality of objects is allowed to execute mutual exclusion operation;

determining a set of objects, of which the number of objects included in the plurality of sets of objects is greater than 1, as the first set of objects.

13. A pushing device for pushing an object to be pushed is characterized by comprising:

14. A storage medium, characterized in that the storage medium comprises a stored program, wherein the program when executed performs the method of any of the preceding claims 1 to 12.

15. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the method of any of the preceding claims 1 to 12 by means of the computer program.