CN116957676A

CN116957676A - Object management method, device, electronic equipment and storage medium

Info

Publication number: CN116957676A
Application number: CN202310125223.XA
Authority: CN
Inventors: 林炳怀; 王丽园; 李昊宇
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2023-02-02
Filing date: 2023-02-02
Publication date: 2023-10-27

Abstract

The application provides an object management method, an object management device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring behavior data of the marked object corresponding to a plurality of operation types of the appointed multimedia information; the behavior data of each operation type is used for representing behavior data generated by the marked object for performing corresponding type operation on the specified multimedia information; based on behavior data of a plurality of operation types, predicting the behavior probability of an object to be managed on the operation types aiming at the appointed multimedia information to obtain the behavior probability corresponding to the operation types; and extracting the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain a target management object. The technical scheme of the embodiment of the application can improve the accuracy of acquiring the target management object.

Description

Object management method, device, electronic equipment and storage medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to an object management method, an object management device, an electronic device, and a storage medium.

Background

The screening of the target management object of the multimedia information can improve the efficiency of multimedia information delivery, at present, the determination of the target management object mostly carries out object expansion through a small amount of tag data (data of the object which historically receives the multimedia information) and an algorithm evaluation model, such as a model based on similarity, characteristic representation learning is carried out on the object to be managed, and a certain measurement mode (cosine similarity, euclidean distance and the like) is used for calculating the characteristic distance between the objects, so that the object is similar to the object which historically receives the multimedia information.

However, the scheme of expanding the object by the algorithm evaluation model is that the label data are limited, the obtained target management object is not high in accuracy, the multimedia information cannot be accurately attached, and when the characteristic distance is calculated, the calculated amount is large, the calculated cost is high, and the waste of calculation resources is easily caused.

Disclosure of Invention

To solve the above technical problems, embodiments of the present application provide an object management method and apparatus, an electronic device, a computer readable storage medium, and a computer program product.

According to an aspect of an embodiment of the present application, there is provided an object management method including: acquiring behavior data of the marked object corresponding to a plurality of operation types of the appointed multimedia information; the behavior data of each operation type is used for representing behavior data generated by the marked object for performing corresponding type operation on the appointed multimedia information; based on the behavior data of the operation types, predicting the behavior probability of the object to be managed on the operation types aiming at the appointed multimedia information to obtain the behavior probability corresponding to the operation types; and extracting the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain a target management object.

According to an aspect of an embodiment of the present application, there is provided an object management apparatus including: the behavior data acquisition module is configured to acquire behavior data of a marked object corresponding to a plurality of operation types of the appointed multimedia information; the behavior data of each operation type is used for representing behavior data generated by the marked object for performing corresponding type operation on the appointed multimedia information; the behavior probability acquisition module is configured to predict the behavior probability of the object to be managed on the operation types aiming at the appointed multimedia information based on the behavior data of the operation types, so as to obtain the behavior probabilities corresponding to the operation types; the target management object acquisition module is configured to extract the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain a target management object.

In one embodiment, the behavior probability acquisition module includes:

a feature data acquisition unit configured to acquire feature data of the object to be managed; wherein the characteristic data is data related to the attribute of the object to be managed;

A sample determination unit configured to take behavior data of a plurality of operation types of the marked object as positive samples and the operation data as negative samples;

and the behavior probability determining unit is configured to predict the behavior probabilities of the object to be managed on the operation types aiming at the appointed multimedia information through the positive sample and the negative sample, so as to obtain the behavior probabilities corresponding to the operation types.

In an embodiment, the behavior probability determination unit includes:

the first behavior probability acquisition plate is configured to input the positive sample and the negative sample into a first model trained in advance to predict the touch operation type, so as to obtain the behavior probability of the object to be managed for the touch operation type;

the second behavior probability plate is configured to input the positive sample and the negative sample into a pre-trained second model to predict the type of the download conversion operation, so as to obtain the behavior probability of the object to be managed for the type of the download conversion operation; and

and the third behavior probability plate is configured to input the positive sample and the negative sample into a pre-trained third model to predict the payment conversion operation type, so as to obtain the behavior probability of the object to be managed for the payment conversion operation type.

In an embodiment, the second behavior probability plate comprises:

the first initial behavior probability obtaining sub-layout is configured to input the positive sample and the negative sample into the second model to obtain initial behavior probability of the object to be managed, which is output by the second model, for a download conversion operation type;

the second behavior probability obtaining sub-layout is configured to calculate the behavior probability of the object to be managed for the download conversion operation type based on the behavior probability of the object to be managed for the touch operation type and the initial behavior probability of the object to be managed for the download conversion operation type.

In an embodiment, the third behavior probability plate comprises:

a third initial behavior probability obtaining sub-block configured to input the positive sample and the negative sample into the third model to obtain initial behavior probability of the object to be managed, which is output by the third model, for a payment conversion operation type;

the third behavior probability obtaining sub-layout is configured to calculate the behavior probability of the object to be managed for the payment conversion operation type based on the initial behavior probability of the object to be managed for the payment conversion operation type, the behavior probability of the object to be managed for the download conversion operation type and the behavior probability of the object to be managed for the touch operation type.

In one embodiment, the object management apparatus further includes:

the training data acquisition module is configured to acquire behavior data of training the marked object; the training marked object is an object of a touch operation type aiming at the appointed multimedia information;

the first training module is configured to train the initial second model by taking the behavior data of the conversion control object with the download conversion operation type in the training control objects as positive samples and taking the behavior data of the object without the download conversion operation type in the marked objects as negative samples to obtain the pre-trained second model;

and the second training module is configured to train the initial third model by taking the behavior data of the object with the payment conversion operation type in the conversion comparison object as a positive sample and taking the behavior data of the object without the payment conversion operation type in the conversion comparison object as a negative sample to obtain the pre-trained third model.

In an embodiment, the plurality of operation types includes a payment conversion operation type, and the target management object acquisition module includes:

an initial probability threshold value obtaining unit configured to allocate initial probability threshold values for the plurality of operation types, respectively;

The to-be-processed object acquisition unit is configured to extract to-be-managed objects with the behavior probability corresponding to each operation type being greater than the corresponding initial probability threshold value, so as to obtain a plurality of to-be-processed objects;

the behavior probability threshold determining unit is configured to perform optimization processing on the initial probability threshold according to the payment conversion value corresponding to the to-be-processed object of the payment conversion operation types in the plurality of to-be-processed objects to obtain a behavior probability threshold corresponding to each operation type;

and the target management object acquisition unit is configured to extract the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain the target management object.

In an embodiment, the behavior probability threshold determining unit comprises:

an average value plate configured to calculate an average payment conversion value of the plurality of objects to be processed according to the payment conversion value corresponding to the object to be processed of the type of payment conversion operation performed on the plurality of objects to be processed;

the conversion duty ratio plate is configured to acquire the duty ratio of the object to be processed of the payment conversion operation type in the plurality of objects to be processed, and obtain the duty ratio of the payment conversion object;

And the behavior probability threshold determining plate is configured to perform operation optimization on the initial probability threshold based on the average payment conversion value and the payment conversion object occupation ratio to obtain a behavior probability threshold corresponding to each operation type, so that the number of objects to be processed, of which the behavior probability corresponding to each operation type is larger than the corresponding behavior probability threshold, meets the preset number.

According to an aspect of an embodiment of the present application, there is provided an electronic device including one or more processors; and storage means for storing one or more computer programs which, when executed by the one or more processors, cause the electronic device to implement the object management method as described above.

According to an aspect of an embodiment of the present application, there is provided a computer-readable storage medium having stored thereon computer-readable instructions, which when executed by a processor of a computer, cause the computer to perform the object management method as described above.

According to an aspect of embodiments of the present application, there is provided a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the object management method provided in the above-described various alternative embodiments.

According to an aspect of an embodiment of the present application, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the steps of an object management method as described above.

In the technical scheme provided by the embodiment of the application, the behavior probability of the object to be managed on different operation types is predicted through the behavior data of the marked object, so that the similarity of the object to be managed is not only simply predicted, but the probability of the object to be managed on executing different operation types aiming at the appointed multimedia information is predicted, the target management object with higher probability of executing different operation types aiming at the appointed multimedia information is extracted by combining the behavior probability threshold values of different operation types, the target management object which accords with the appointed multimedia information is obtained, and the screening accuracy of the target management object is improved; on the other hand, the object management method does not need to determine the target management object by calculating the characteristic distance, can greatly reduce the calculated amount of object management and liberate the calculation resources.

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 application as claimed.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application. It is evident that the drawings in the following description are only some embodiments of the present application and that other drawings may be obtained from these drawings without inventive effort for a person of ordinary skill in the art. In the drawings:

FIG. 1 is a schematic illustration of an implementation environment in which the present application is directed;

FIG. 2 is a flow chart of an object management method shown in an exemplary embodiment of the application;

FIG. 3 is a schematic diagram of an operation type structure shown in an exemplary embodiment of the present application;

FIG. 4 is a flow chart of step S230 in the embodiment of FIG. 2 in an exemplary embodiment;

FIG. 5 is a flow chart of step S450 in the embodiment of FIG. 4 in an exemplary embodiment;

FIG. 6 is a schematic diagram illustrating output relationships between models of different types of operations according to an exemplary embodiment of the present application;

FIG. 7 is a flow chart of step S250 in the embodiment of FIG. 2 in an exemplary embodiment;

FIG. 8 is a flow chart of an object management method shown in another exemplary embodiment of the application;

FIG. 9 is a schematic diagram illustrating an interface for uploading tagged objects, according to an illustrative embodiment of the application;

FIG. 10 is an interface diagram of a download target management object shown in an exemplary embodiment of the present application;

fig. 11 is a schematic structural view of an object management apparatus according to an exemplary embodiment of the present application;

fig. 12 is a schematic diagram of a computer system suitable for use in implementing an embodiment of the application.

Detailed Description

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the application. Rather, they are merely examples of apparatus and methods consistent with aspects of the application as detailed in the accompanying claims.

The block diagrams depicted in the figures are merely functional entities and do not necessarily correspond to physically separate entities. That is, the functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor devices and/or microcontroller devices.

The flow diagrams depicted in the figures are exemplary only, and do not necessarily include all of the elements and operations/steps, nor must they be performed in the order described. For example, some operations/steps may be decomposed, and some operations/steps may be combined or partially combined, so that the order of actual execution may be changed according to actual situations.

Also to be described is: in the present application, the term "plurality" means two or more. "and/or" describes an association relationship of an association object, meaning that there may be three relationships, e.g., a and/or B may represent: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship.

It will be appreciated that in the specific embodiment of the present application, information such as behavior data, feature data, etc. of the object is related, and when the above embodiment of the present application applies the information to a specific product or technology, it is required to obtain permission or consent of the user, or perform related desensitization filtering processing, and the collection, use and processing of related information is required to comply with related laws and regulations and standards of related countries and regions.

Description of technical words:

cross entropy loss: in the information theory, cross entropy is the difference representing two probability distributions (true distribution and predicted distribution).

Operation optimization: the optimization problem (Optimization problem) is in the field of mathematics and computer science, the problem of finding the best solution from all possible solutions. Optimization problems can be divided into two categories, depending on whether the variables are continuous or discrete: continuous optimization problem and combinatorial optimization.

An optimization solver: an algorithm package that solves the integer programming model specifically.

The object is: the object in this embodiment is regarded as a user who can receive multimedia information using a terminal and perform related operations.

Multimedia information: advertising information in the form of video, text, pictures, links, etc.

Optimized distributed gradient enhancement library (XGBoost): a general Tree Boosting algorithm, one representative of which is a gradient Boosting decision Tree (Gradient Boosting Decision Tree, GBDT), also known as MART (Multiple Additive Regression Tree).

An optimization solver: the business problem is mathematically modeled and solved to help generate a logic-compliant accurate decision. Based on the optimization solver, decision optimization can be realized, so that the efficiency is improved, the cost is reduced, and the profitability is improved. General optimization solvers provide flexible high-performance mathematical programming solvers for linear programming, mixed integer programming, quadratic programming, and quadratic constraint programming problems, including distributed parallel algorithms for mixed integer programming, supporting the use of multiple computers to solve problems. Such as CPLEX solver and SCIP solver.

Artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use the knowledge to obtain optimal results. In other words, artificial intelligence is an integrated technology of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence, i.e. research on design principles and implementation methods of various intelligent machines, enables the machines to have functions of sensing, reasoning and decision.

The artificial intelligence technology is a comprehensive subject, and relates to the technology with wide fields, namely the technology with a hardware level and the technology with a software level. Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.

Machine Learning (ML) is a multi-domain interdisciplinary, involving multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, etc. It is specially studied how a computer simulates or implements learning behavior of a human to acquire new knowledge or skills, and reorganizes existing knowledge structures to continuously improve own performance. Machine learning is the core of artificial intelligence, a fundamental approach to letting computers have intelligence, which is applied throughout various areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, confidence networks, reinforcement learning, transfer learning, induction learning, teaching learning, and the like.

At present, screening of target management objects of multimedia information is performed based on a small amount of tag data, namely similar object expansion is performed based on data of objects which historically accept the multimedia information. Specifically, a technique of evaluating a model by an algorithm based on a small set of objects that have historically accepted the multimedia information, and then finding more potentially relevant similar objects from among a vast number of candidate objects. There are two main types of similar object expansion solutions at present: (1) a classification-based model; (2) similarity-based models.

Based on the classified model, the object which receives the multimedia information historically is taken as a positive sample, a negative sample is screened from a large number of candidate objects, two classifications are carried out, a probability threshold is set, model classification scoring is carried out on the large number of candidate objects, and an object list meeting the requirements is screened based on the threshold to serve as a target management object.

And based on the similarity model, mainly encoding each object in the objects which historically accept the multimedia information, wherein the encoding can be from a feature representation model trained based on the full-scale object, and acquiring the feature representation of each object, namely user embedding. Each history object passes through a characteristic representation model to obtain characteristic representation of each object in the objects which receive the multimedia information in history, characteristic representation learning is carried out on candidate objects, a certain measurement mode (cosine similarity, euclidean distance and the like) is used for calculating the distance between users and enabling a batch of objects which are closest to the objects which receive the multimedia information in history to be directly returned as an expansion result.

In the existing scheme, because object information of the multimedia information is limited in history acceptance, a more effective target management object cannot be constructed based on a two-class model of the information so as to meet the requirement of multimedia information delivery, a better representation model is needed based on an ebedding mode, and meanwhile, in the process of similarity comparison, the calculation amount is large, and the cost is high.

Referring first to fig. 1, fig. 1 is a schematic diagram of an implementation environment according to the present application. The implementation environment includes a terminal 100 and a server side 200, and communication is performed between the terminal 100 and the server side 200 through a wired or wireless network.

Of course, the number of server sides 200 in fig. 1 is merely exemplary, and in other embodiments, other numbers of server sides 200 are also possible, in this embodiment, the terminal 100 may be configured to determine an object to be managed, that is, an object to be managed, and receive behavior data of a marked object for specifying multiple operation types corresponding to multimedia information, where the behavior data may include attribute information and behavior information of the marked object, the attribute information may be information such as height, age, work, address, etc., and the behavior information is information of an operation type performed for the multimedia information, such as information related to a touch operation type, a download conversion operation type, and a payment conversion operation type for the multimedia information.

The terminal 100 also sends the object to be managed and the behavior data of the marked object to the server 200, so that the server 200 performs object management based on the object to be managed and the behavior data of the marked object to obtain an object management result, and the server 200 returns the object management result to the terminal 100 and displays the object management result through a visualization module carried by the terminal 100.

Of course, the object to be managed may be a full-scale object, that is, the object management is to screen similar objects in the object to be managed based on the behavior data of the marked object corresponding to the specified multimedia information, so in some embodiments, the object to be managed is stored by the server 200 itself, that is, the terminal 100 may directly send the behavior data of the marked object to the server 200.

For example, after obtaining the object to be managed and the behavior data of the marked object, the terminal 100 sends the data of the object to the server 200, and the server 200 predicts the behavior probability of the object to be managed on the multiple operation types for the specified multimedia information based on the behavior data of the multiple operation types, to obtain the behavior probabilities corresponding to the multiple operation types; based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type, extracting the object to be managed to obtain a target management object, and returning the target management object to the terminal 100 for display through a visualization module of the terminal 100.

The terminal 100 may be a mobile phone, a computer, an intelligent voice interaction device, an intelligent home appliance, a vehicle-mounted terminal, an aircraft, etc., which is not limited herein. The server 200 may be an independent physical server, or may be a server cluster or a distributed system formed by a plurality of physical servers, where a plurality of servers may form a blockchain, and the servers are nodes on the blockchain, and the server 200 may also be a cloud server that provides 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), and basic cloud computing services such as big data and artificial intelligent platforms, which are not limited herein.

It should be noted that this embodiment is only an exemplary implementation environment provided for the convenience of understanding the idea of the present application, and should not be construed as providing any limitation on the scope of use of the present application.

Fig. 2 is a flowchart illustrating an object management method according to an exemplary embodiment, which may be applied to the implementation environment in fig. 1 and executed by the server side 200 in fig. 1, it should be understood that the method may also be applied to other exemplary implementation environments and executed by devices in other implementation environments, and the embodiment is not limited to the implementation environment to which the method is applied.

In an exemplary embodiment, the method may include steps S210 to S250, which are described in detail as follows:

step S210: behavior data of the marked object for a plurality of operation types corresponding to the specified multimedia information is obtained.

In this embodiment, the marked object is an object that has performed a related operation on the specified multimedia information or the multimedia information of the same domain/same type/similar category as the specified multimedia information, where the specified multimedia information may be certain specific advertisement information or certain specific advertisement information.

The behavior data of each operation type is used for characterizing the behavior data generated by the marked object for performing the corresponding type of operation on the specified multimedia information, and the behavior data also comprises the basic attribute information of the specified object.

For the operations of specifying multimedia information, the delivering party of the specified multimedia information is not only satisfied with the click or conversion rate, but also pays more attention to post-effect conversion, so the multiple operation types in this embodiment can be roughly classified into three types, such as a touch operation type, a download conversion operation type and a payment conversion operation type, and the structures of the three operation types can be regarded as funneling, as shown in fig. 3, as the target approaches to the payment conversion operation type, the number of payment conversion operation types performed in the marked object is smaller and smaller.

The touch operation type can be regarded as clicking the multimedia information for viewing, the download conversion operation type can be regarded as carrying out download registration on the application corresponding to the multimedia information, and the payment conversion operation type can be regarded as carrying out payment use on the application corresponding to the multimedia information.

If in the game scene, the touch operation type refers to clicking the multimedia information of the game scene, the download conversion operation type refers to creating a role, and the payment conversion operation type refers to how much money the game player pays.

Step S230: based on the behavior data of the operation types, predicting the behavior probability of the object to be managed on the operation types aiming at the appointed multimedia information, and obtaining the behavior probability corresponding to the operation types.

In this embodiment, the object to be managed may be regarded as a full-scale object, or may be an object obtained by randomly extracting the full-scale object.

For different operation types, different models may be constructed, so that each operation type may obtain a model, which may be a machine network model for predicting probability, such as XGBoost (optimized distributed gradient enhancement library), without specific limitation.

In this way, the feature data of the object to be managed and the behavior data of the marked object in a plurality of operation types are respectively input into the models of different operation types, the feature data of the object to be managed is used as a negative sample, the behavior data of the marked object in a plurality of operation types is used as a positive sample, and finally the behavior probability of the object to be managed corresponding to the operation types of the appointed multimedia information can be obtained.

In this embodiment, the feature data is data related to the attribute of the object to be managed, such as attribute information of the object to be managed and behavior data generated by corresponding type of operation performed by the object to be managed on other multimedia information.

That is, finally, the object to be managed may obtain respective behavior probabilities for the touch operation type, the download conversion operation type, and the payment conversion operation type.

Step S250: and extracting the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain a target management object.

In this embodiment, the target management object is determined by the behavior probability corresponding to each operation type of the object to be managed, specifically, a corresponding behavior probability threshold may be determined for each operation type, and in the object to be managed, an object whose behavior probability of each operation type is greater than the corresponding behavior probability threshold is the target management object, and of course, the target management objects obtained by different operation types may overlap, and at this time, deduplication processing is performed, so as to obtain the final target management object.

The action probability threshold corresponding to each operation type can be determined through operation optimization, namely, the object of the operation type for payment conversion in the target management object is the highest in duty ratio, and the average payment conversion value is the largest.

In this embodiment, the behavior probability of the object to be managed on different operation types is predicted by the behavior data of the marked object, so that, based on different operation types, the target management object with higher conversion rate for the specified multimedia information is extracted by combining the behavior probability thresholds of different operation types, thereby obtaining the target management object which accords with the specified multimedia information delivery, and improving the accuracy of the screening of the target management object.

In this embodiment, requirements for multimedia information delivery are more and more complex, such as not only meeting the requirements of clicking advertisements, but also paying attention to transformation behaviors (such as payment behaviors) after clicking advertisements, fully considering requirements of complex scenes, constructing a plurality of models, constructing positive and negative sample sets through behavior data of marked objects and feature data of objects to be managed, and predicting behavior probabilities corresponding to a plurality of operation types.

On the other hand, in the embodiment, the target management object is determined by the behavior probabilities of the objects to be managed on different operation types, so that the feature distance between the objects is not required to be calculated, the calculation amount of object management can be greatly reduced, and the calculation resources are reduced.

Fig. 4 is a flowchart of step S230 in an exemplary embodiment in the embodiment shown in fig. 2. As shown in fig. 4, in an exemplary embodiment, step S230 predicts the behavior probabilities of the object to be managed on the plurality of operation types for the specified multimedia information based on the behavior data of the plurality of operation types, and the process of obtaining the behavior probabilities corresponding to the plurality of operation types may include steps S410 to S450, which are described in detail below:

step S410: and acquiring characteristic data of the object to be managed.

In this embodiment, the feature data is data related to the attribute of the object to be managed, such as behavior data for a plurality of operation types corresponding to the multimedia information of the same type as the specified multimedia information, and attribute information of the object to be managed.

Step S430: behavior data of a plurality of operation types of the marked object is taken as a positive sample, and operation data is taken as a negative sample.

In this embodiment, feature construction may be performed based on behavior data of multiple operation types of the marked object and feature data of the object to be managed, corresponding feature production may be performed from a feature library, operations such as searching, feature selection, feature encoding, feature dimension reduction, etc. may be performed on the object feature based on the original feature library, and finally, a feature library of the marked object and a feature library of the object to be managed may be obtained, and at the same time, the feature library of the marked object may be used as a positive sample, and the feature library of the object to be managed may be used as a negative sample to perform behavior probability prediction.

Step S450: and predicting the behavior probabilities of the object to be managed on a plurality of operation types aiming at the appointed multimedia information through the positive sample and the negative sample to obtain the behavior probabilities corresponding to the operation types.

In this embodiment, different models exist for different operation types, so that a positive sample and a negative sample can be respectively input into the models of multiple operation types, and the behavior probability of an object to be managed on the multiple operation types for the specified multimedia information is predicted, so as to obtain the behavior probabilities corresponding to the multiple operation types.

In this embodiment, feature extraction is performed on the data of the marked object and the data of the object to be managed, so that prediction calculation is performed through models of different operation types, so as to obtain the behavior probabilities of the object to be managed on a plurality of operation types aiming at the specified multimedia information, and then the target management object can be accurately screened through the behavior probabilities on the plurality of operation types.

Fig. 5 is a flowchart of step S450 in an exemplary embodiment of the embodiment shown in fig. 4. As shown in fig. 5, in an exemplary embodiment, step S450 predicts the behavior probabilities of the object to be managed on the plurality of operation types for the specified multimedia information through the positive sample and the negative sample, and the process of obtaining the behavior probabilities corresponding to the plurality of operation types may include steps S510 to S550, which are described in detail as follows:

step S510: and inputting the positive sample and the negative sample into a pre-trained first model to predict the touch operation type, and obtaining the behavior probability of the object to be managed for the touch operation type.

In this embodiment, three models are respectively constructed for the touch operation type, the download conversion operation type and the payment conversion operation type, namely, the first model, the second model and the third model are correspondingly used, the first model, the second model and the third model are trained in advance, and after the training is finished, the three models can be used for determining the behavior probabilities on a plurality of operation types.

In one embodiment, the training process for the first model is: and training the initial first model to obtain a trained first model by taking behavior data of the trained marked object as a positive sample and characteristic data of the object to be managed as a negative sample, wherein the trained marked object is an object of a touch operation type aiming at the appointed multimedia information.

Then, the first model is used for predicting and obtaining the behavior probability of the object to be managed for the touch operation type, specifically, the positive sample and the negative sample are input into the first model trained in advance, and the first model outputs the behavior probability of the object to be managed for the touch operation type.

Step S530: and inputting the positive sample and the negative sample into a pre-trained second model to predict the type of the download conversion operation, and obtaining the behavior probability of the object to be managed for the type of the download conversion operation.

In one embodiment, the training process for the second model is: and training the initial second model by taking the behavior data of the conversion control object with the download conversion operation type in the training control object as a positive sample and the behavior data of the object without the download conversion operation type in the marked object as a negative sample to obtain the second model.

Since the behavior probabilities of different operation types for formulating the multimedia information need to be predicted, and the download conversion operation type must occur after the object performs the touch operation type, the probability value output by the second model is not the final behavior probability of the object to be managed for the download conversion operation type.

In a specific embodiment, referring to fig. 6, a schematic diagram of output relationships between models of different types of operations may be shown, where a positive sample and a negative sample are input into a second model, so as to obtain an initial behavior probability of an object to be managed for a download conversion operation type output by the second model, and the obtained product is the behavior probability of the object to be managed for the download conversion operation type only by multiplying the behavior probability of the object to be managed for the touch operation type and the initial behavior probability of the object to be managed for the download conversion operation type.

Specifically, the behavior probability of the object to be managed for the touch operation type may be regarded as P touch |to be managed), the initial behavior probability of the object to be managed output by the second model for the download conversion operation type may be regarded as P download|touch), the behavior probability of the object to be managed for the download conversion operation type is P download|to be managed), and the manner of obtaining P download|to be managed) should be:

p download |to be managed) =p (touch |to be managed) ×p (download|touch).

Step S550: and inputting the positive sample and the negative sample into a pre-trained third model to predict the payment conversion operation type, and obtaining the behavior probability of the object to be managed for the payment conversion operation type.

In one embodiment, the training process for the third model is: and taking the behavior data of the object with the payment conversion operation type in the conversion comparison object as a positive sample, taking the behavior data of the object without the payment conversion operation type in the conversion comparison object as a negative sample, and training the initial third model to obtain the third model.

Likewise, the payment conversion operation type must occur after the object performs the download conversion operation type, and thus, the probability value output by the third model is not the final behavior probability of the object to be managed for the payment conversion operation type.

And inputting the positive sample and the negative sample into a second model to obtain initial behavior probability of the object to be managed, which is output by the second model, for the download conversion operation type, and multiplying the behavior probability of the object to be managed for the touch operation type and the initial behavior probability of the object to be managed for the download conversion operation type to obtain the product which is the behavior probability of the object to be managed for the download conversion operation type.

Referring to fig. 6, in a specific embodiment, the initial behavior probability for the download transformation operation type may be P pays |download), and the behavior probability for the download transformation operation type P pays|to be managed) is calculated as follows:

P payment |to be managed) =p (touch |to be managed) ×p (download |touch) ×p (payment|download).

In the embodiment, by setting a plurality of models, the behavior probability of the object to be managed on a plurality of operation types is obtained through analysis, so that the screening of the target management object of the complex scene of multimedia information delivery is met, and the effectiveness of the multimedia information delivery is improved.

Fig. 7 is a flowchart of step S250 in an exemplary embodiment of the embodiment shown in fig. 2. As shown in fig. 7, in an exemplary embodiment, step S250 extracts an object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type, and the process of obtaining the target management object may include steps S710 to S770, which are described in detail below:

step S710: an initial probability threshold is assigned to each of the plurality of operation types.

In this embodiment, the determination of the target management object may be implemented by three probabilities of the object to be managed, specifically, an initial probability threshold is allocated to each of the multiple operation types, that is, an initial probability threshold t1 of the touch operation type, an initial probability threshold t2 of the download conversion operation type, and an initial probability threshold t3 of the payment conversion operation type.

The range of t1, t2 and t3 which can be valued is limited, a click model is taken as an example, the probability of clicking the model in a test set is ordered, the probability is ordered from high to low, 100 bins are carried out, the enumerated value of t1 is 100, and finally, the goal of the whole integer programming is to enable the average charge of people of a target management object to be highest, and the charge proportion is highest.

Step S730: and extracting the objects to be managed, of which the behavior probability corresponding to each operation type is larger than the corresponding initial probability threshold, and obtaining a plurality of objects to be processed.

In this embodiment, for the behavior probability corresponding to each operation type of the object to be managed, the behavior probabilities of different operation types are compared with the behavior probability threshold of the corresponding operation type, and the object to be managed, whose behavior probability corresponding to each operation type is greater than the corresponding initial probability threshold, is used as the object to be processed.

The object to be processed is an object with a behavior probability larger than t1 for the touch operation type in the object to be managed, an object with a behavior probability larger than t2 for the download conversion operation type in the object to be managed, and an object with a behavior probability larger than t3 for the payment conversion operation type in the object to be managed.

Of course, there are cases where the behavior probability threshold of a certain object to be managed on multiple operation types is greater than the corresponding initial behavior probability threshold, and in this case, the object to be processed may be subjected to deduplication first.

Step S750: and optimizing the initial probability threshold according to the payment conversion value corresponding to the object to be processed of the payment conversion operation type in the plurality of objects to be processed, so as to obtain the behavior probability threshold corresponding to each operation type.

In this embodiment, since the feature data of the object to be managed includes behavior data of a plurality of operation types corresponding to other multimedia information of the same category as the specified multimedia information, that is, data related to performing a payment conversion operation type with respect to other multimedia information of the same category as the specified multimedia information, a payment conversion value corresponding to the object to be processed, which performs the payment conversion operation type with respect to other multimedia information, in the object to be processed may be extracted from the feature data.

Then, according to the payment conversion value corresponding to the object to be processed of the payment conversion operation type in the plurality of objects to be processed, calculating an average payment conversion value of the plurality of objects to be processed, wherein the calculation mode of the average payment conversion value C is as follows:

Where n is the number of objects to be processed, i is the ith object to be processed, c _i And (3) carrying out payment conversion values corresponding to the to-be-processed objects of the payment conversion operation type in the ith to-be-processed object.

Meanwhile, the duty ratio of the objects to be processed of the types of the payment conversion operation in the objects to be processed can be obtained based on the payment conversion value corresponding to the objects to be processed of the types of the payment conversion operation in the objects to be processed, and the duty ratio of the objects to be processed is obtained, wherein the duty ratio of the objects to be processed is the number of the objects to be processed of the types of the payment conversion operation in the objects to be processed/the total number of the objects to be processed.

The payment conversion object duty ratio can be regarded as a payment rate, and aiming at the delivery of the specified multimedia information, the required maximum target is the payment rate and the average payment conversion value is highest, so that an initial probability threshold can be respectively allocated to a plurality of operation types based on an operation preparation optimization algorithm, and a final behavior probability threshold corresponding to each operation type is obtained.

In a specific embodiment, the objectives that may be set for operational optimization are:

y＝maximize(w1*C+w2*R)

wherein w1 and w2 are weights of an average payment conversion value and a payment conversion object ratio respectively, the weight value can be set by experience parameters, the convergence condition of operation optimization is that min is less than or equal to the number of target management objects and less than or equal to max, wherein min and max are the number range of the preset target management objects, namely the finally obtained behavior probability threshold value corresponding to each operation type, so that the number of objects to be processed, of which the behavior probability corresponding to each operation type is greater than the corresponding behavior probability threshold value, meets the preset number.

In the operation optimization target, the size of C is determined by t1, t2 and t3, so that the final t1, t2 and t3, namely the final behavior probability threshold corresponding to each operation type, can be obtained through the operation optimization target and the convergence condition.

In a specific embodiment, implementation of operational optimization is accomplished by an optimization solver.

Step S770: and extracting the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain a target management object.

In this embodiment, objects to be managed in each operation type with a behavior probability greater than a corresponding behavior probability threshold are screened, and then, deduplication processing is performed to obtain a final target management object.

In this embodiment, based on operation planning optimization, multi-objective decision is performed, that is, based on behavior data of marked objects, an optimization constraint problem is constructed, based on an optimization solver, the most available strategy is solved, and the most available strategy is used as a delivery strategy of appointed multimedia information, so that the efficiency of appointed multimedia information delivery is improved.

Fig. 8 is a flowchart of an embodiment of an object management method, specifically, an object management device corresponding to the object management method is disposed in a server, a terminal connected by a designated multimedia information delivering party through the server is uploaded with a marked object and behavior data of the marked object at an interface shown in fig. 9, the server transmits the marked object to the object management device to perform object management shown in fig. 2 to 7, then, a target management object is generated in the object management device, and is displayed at the terminal interface shown in fig. 10 through the server, and the designated multimedia information delivering party can download the target management object through the terminal interface shown in fig. 10 to perform delivery of the designated multimedia information.

In a specific embodiment, the specified multimedia information is an advertisement of a certain automobile manufacturer, and the delivery efficiency a/B test is performed by the object management method in the embodiment, the object management method based on the classified model and the object management method based on the similarity model to obtain the test result shown in table 1, where the conversion rate is the cost of purchasing goods corresponding to the advertisement of the certain automobile manufacturer by performing advertisement of the certain automobile manufacturer on the target management object obtained by different object management methods, and the conversion cost is the cost of obtaining the target management object by different object management methods.

Object management mode	Conversion rate	Conversion cost
			This embodiment	0.81％	32.78
Classification-based model	0.66％	39.55
			Similarity-based model	0.74％	56.90

TABLE 1

As can be seen from table 1, the object management method in this embodiment has great advantages in terms of advertisement conversion rate and average conversion cost.

Fig. 11 is a schematic structural view of an object management apparatus according to an exemplary embodiment. As shown in fig. 11, in an exemplary embodiment, the object management apparatus includes:

a behavior data acquisition module 1110 configured to acquire behavior data of a plurality of operation types corresponding to the marked object for the specified multimedia information; the behavior data of each operation type is used for representing behavior data generated by the marked object for performing corresponding type operation on the specified multimedia information;

The behavior probability obtaining module 1130 is configured to predict the behavior probabilities of the object to be managed on the multiple operation types for the specified multimedia information based on the behavior data of the multiple operation types, so as to obtain the behavior probabilities corresponding to the multiple operation types;

the target management object obtaining module 1150 is configured to extract the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type, and obtain the target management object.

The object management device in the embodiment can manage the object and improve the accuracy of the appointed multimedia information.

In one embodiment, the behavior probability acquisition module includes:

a feature data acquisition unit configured to acquire feature data of an object to be managed; wherein, the characteristic data is data related to the attribute of the object to be managed;

and the behavior probability determining unit is configured to predict the behavior probabilities of the object to be managed on a plurality of operation types aiming at the appointed multimedia information through the positive sample and the negative sample, so as to obtain the behavior probabilities corresponding to the operation types.

In an embodiment, the behavior probability determination unit includes:

the first behavior probability acquisition plate is configured to input a positive sample and a negative sample into a pre-trained first model to predict a touch operation type, so as to obtain the behavior probability of an object to be managed for the touch operation type;

In an embodiment, the second behavior probability plate comprises:

the first initial behavior probability obtaining sub-layout is configured to input a positive sample and a negative sample into the second model to obtain initial behavior probability of the object to be managed, which is output by the second model, for the download conversion operation type;

In an embodiment, the third behavior probability plate comprises:

the third initial behavior probability obtaining sub-layout is configured to input a positive sample and a negative sample into the third model to obtain initial behavior probability of the object to be managed, which is output by the third model, for the payment conversion operation type;

In one embodiment, the object management apparatus further includes:

the training data acquisition module is configured to acquire behavior data of training the marked object; training the marked object as an object of a touch operation type aiming at the appointed multimedia information;

the first training module is configured to train the initial second model by taking the behavior data of the conversion control object of the download conversion operation type in the training control object as a positive sample and the behavior data of the object of the non-download conversion operation type in the marked object as a negative sample to obtain a pre-trained second model;

And the second training module is configured to train the initial third model by taking the behavior data of the object with the payment conversion operation type in the conversion comparison object as a positive sample and taking the behavior data of the object without the payment conversion operation type in the conversion comparison object as a negative sample to obtain a pre-trained third model.

In one embodiment, the plurality of operation types includes a payment conversion operation type, and the target management object acquisition module includes:

an initial probability threshold value acquisition unit configured to allocate initial probability threshold values for the plurality of operation types, respectively;

the behavior probability threshold determining unit is configured to optimize the initial probability threshold according to the payment conversion value corresponding to the to-be-processed object of the payment conversion operation type in the plurality of to-be-processed objects to obtain a behavior probability threshold corresponding to each operation type;

the target management object acquisition unit is configured to extract the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain the target management object.

the average value plate is configured to calculate average payment conversion values of the plurality of objects to be processed according to the payment conversion values corresponding to the objects to be processed of the payment conversion operation type in the plurality of objects to be processed;

the behavior probability threshold determining plate is configured to perform operation optimization on the initial probability threshold based on the average payment conversion value and the payment conversion object occupation ratio to obtain a behavior probability threshold corresponding to each operation type, so that the number of objects to be processed, of which the behavior probability corresponding to each operation type is larger than the corresponding behavior probability threshold, meets the preset number.

It should be noted that, the object management apparatus provided in the foregoing embodiment and the object management method provided in the foregoing embodiment belong to the same concept, and a specific manner in which each module and unit perform an operation has been described in detail in the method embodiment, which is not described herein again.

The embodiment of the application also provides electronic equipment, which comprises: one or more processors; and a storage means for storing one or more programs which, when executed by the one or more processors, cause the electronic device to implement the object management method provided in the respective embodiments described above.

Fig. 12 shows a schematic diagram of a computer system suitable for use in implementing an embodiment of the application.

It should be noted that, the computer system 1200 of the electronic device shown in fig. 12 is only an example, and should not impose any limitation on the functions and the application scope of the embodiments of the present application.

As shown in fig. 12, the computer system 1200 includes a central processing unit (Central Processing Unit, CPU) 1201 which can perform various appropriate actions and processes, such as performing the methods in the above-described embodiments, according to a program stored in a Read-Only Memory (ROM) 1202 or a program loaded from a storage section 1208 into a random access Memory (Random Access Memory, RAM) 1203. In the RAM 1203, various programs and data required for the system operation are also stored. The CPU 1201, ROM 1202, and RAM 1203 are connected to each other through a bus 1204. An Input/Output (I/O) interface 1205 is also connected to bus 1204.

The following components are connected to the I/O interface 1205: an input section 1206 including a keyboard, a mouse, and the like; an output portion 1207 including a Cathode Ray Tube (CRT), a liquid crystal display (Liquid Crystal Display, LCD), and a speaker, etc.; a storage section 1208 including a hard disk or the like; and a communication section 1209 including a network interface card such as a LAN (Local Area Network ) card, a modem, or the like. The communication section 1209 performs communication processing via a network such as the internet. The drive 1210 is also connected to the I/O interface 1205 as needed. A removable medium 1211 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is installed as needed on the drive 1210 so that a computer program read out therefrom is installed into the storage section 1208 as needed.

In particular, according to embodiments of the present application, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present application include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising a computer program for performing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via the communication portion 1209, and/or installed from the removable media 1211. When executed by a Central Processing Unit (CPU) 1201, performs the various functions defined in the system of the present application.

It should be noted that, the computer readable medium shown in the embodiments of the present application may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium may be, for example, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-Only Memory (ROM), an erasable programmable read-Only Memory (Erasable Programmable Read Only Memory, EPROM), flash Memory, an optical fiber, a portable compact disc read-Only Memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present application, however, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with a computer-readable computer program embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. A computer program embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the foregoing.

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. Where each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The units involved in the embodiments of the present application may be implemented by software, or may be implemented by hardware, and the described units may also be provided in a processor. Wherein the names of the units do not constitute a limitation of the units themselves in some cases.

Another aspect of the present application also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements an object management method as before. The computer-readable storage medium may be included in the electronic device described in the above embodiment or may exist alone without being incorporated in the electronic device.

Another aspect of the application also provides a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the object management methods provided in the respective embodiments described above.

The foregoing is merely illustrative of the preferred embodiments of the present application and is not intended to limit the embodiments of the present application, and those skilled in the art can easily make corresponding variations or modifications according to the main concept and spirit of the present application, so that the protection scope of the present application shall be defined by the claims.

Claims

1. An object management method, comprising:

acquiring behavior data of the marked object corresponding to a plurality of operation types of the appointed multimedia information; the behavior data of each operation type is used for representing behavior data generated by the marked object for performing corresponding type operation on the appointed multimedia information;

based on the behavior data of the operation types, predicting the behavior probability of the object to be managed on the operation types aiming at the appointed multimedia information to obtain the behavior probability corresponding to the operation types;

and extracting the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain a target management object.

2. The method according to claim 1, wherein predicting the behavior probabilities of the object to be managed on the plurality of operation types for the specified multimedia information based on the behavior data of the plurality of operation types, to obtain the behavior probabilities corresponding to the plurality of operation types, includes:

acquiring characteristic data of the object to be managed; wherein the characteristic data is data related to the attribute of the object to be managed;

Taking behavior data of a plurality of operation types of the marked object as positive samples and taking the operation data as negative samples;

and predicting the behavior probability of the object to be managed on the operation types aiming at the appointed multimedia information through the positive sample and the negative sample to obtain the behavior probability corresponding to the operation types.

3. The method according to claim 2, wherein predicting, by the positive sample and the negative sample, the behavior probabilities of the object to be managed on the plurality of operation types for the specified multimedia information, to obtain the behavior probabilities corresponding to the plurality of operation types, includes:

inputting the positive sample and the negative sample into a pre-trained first model to predict the touch operation type, so as to obtain the behavior probability of the object to be managed for the touch operation type;

inputting the positive sample and the negative sample into a pre-trained second model to predict the type of the download conversion operation, so as to obtain the behavior probability of the object to be managed for the type of the download conversion operation; and

and inputting the positive sample and the negative sample into a pre-trained third model to predict the payment conversion operation type, so as to obtain the behavior probability of the object to be managed for the payment conversion operation type.

4. A method according to claim 3, wherein said inputting the positive and negative samples into a pre-trained second model for prediction of the type of download transformation operation, resulting in a probability of behavior of the object to be managed for the type of download transformation operation, comprises:

inputting the positive sample and the negative sample into the second model to obtain initial behavior probability of the object to be managed, which is output by the second model, for the type of the download conversion operation;

and calculating the behavior probability of the object to be managed for the download conversion operation type based on the behavior probability of the object to be managed for the touch operation type and the initial behavior probability of the object to be managed for the download conversion operation type.

5. A method according to claim 3, wherein said inputting the positive and negative samples into a pre-trained third model for prediction of payment transformation operation type, resulting in a probability of behavior of the object to be managed for payment transformation operation type, comprises:

inputting the positive sample and the negative sample into the third model to obtain initial behavior probability of the object to be managed, which is output by the third model, for a payment conversion operation type;

And calculating the behavior probability of the object to be managed for the payment conversion operation type based on the initial behavior probability of the object to be managed for the payment conversion operation type, the behavior probability of the object to be managed for the downloading conversion operation type and the behavior probability of the object to be managed for the touch operation type.

6. A method according to claim 3, characterized in that the method further comprises:

acquiring behavior data of a training marked object; the training marked object is an object of a touch operation type aiming at the appointed multimedia information;

taking the behavior data of the conversion control object with the download conversion operation type in the training control object as a positive sample, taking the behavior data of the object without the download conversion operation type in the training marked object as a negative sample, and training an initial second model to obtain the pre-trained second model;

and training the initial third model by taking the behavior data of the object with the payment conversion operation type in the conversion comparison object as a positive sample and taking the behavior data of the object without the payment conversion operation type in the conversion comparison object as a negative sample to obtain the pre-trained third model.

7. The method according to claim 1, wherein the plurality of operation types include payment conversion operation types, the extracting the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain a target management object includes:

assigning an initial probability threshold to each of the plurality of operation types;

extracting objects to be managed, of which the behavior probability corresponding to each operation type is larger than a corresponding initial probability threshold, and obtaining a plurality of objects to be processed;

optimizing the initial probability threshold according to the payment conversion value corresponding to the object to be processed of the payment conversion operation type in the plurality of objects to be processed to obtain a behavior probability threshold corresponding to each operation type;

and extracting the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain the target management object.

8. The method of claim 7, wherein the optimizing the initial probability threshold according to the payment conversion value corresponding to the object to be processed for performing the payment conversion operation type in the plurality of objects to be processed to obtain the behavior probability threshold corresponding to each operation type includes:

Calculating the average payment conversion value of the plurality of objects to be processed according to the payment conversion value corresponding to the object to be processed of the payment conversion operation type in the plurality of objects to be processed;

acquiring the duty ratio of the object to be processed of the type of payment conversion operation in the plurality of objects to be processed, and obtaining the duty ratio of the payment conversion object;

and carrying out operation optimization on the initial probability threshold based on the average payment conversion value and the payment conversion object occupation ratio to obtain a behavior probability threshold corresponding to each operation type, so that the number of objects to be processed, of which the behavior probability corresponding to each operation type is larger than the corresponding behavior probability threshold, meets the preset number.

9. An object management apparatus, comprising:

the behavior data acquisition module is configured to acquire behavior data of a marked object corresponding to a plurality of operation types of the appointed multimedia information; the behavior data of each operation type is used for representing behavior data generated by the marked object for performing corresponding type operation on the appointed multimedia information;

the behavior probability acquisition module is configured to predict the behavior probability of the object to be managed on the operation types aiming at the appointed multimedia information based on the behavior data of the operation types, so as to obtain the behavior probabilities corresponding to the operation types;

The target management object acquisition module is configured to extract the object to be managed based on the behavior probability corresponding to each operation type and the behavior probability threshold corresponding to each operation type to obtain a target management object.

10. An electronic device, comprising:

one or more processors;

storage means for storing one or more computer programs which, when executed by the one or more processors, cause the electronic device to implement the method of any of claims 1-8.

11. A computer readable storage medium having stored thereon computer readable instructions which, when executed by a processor of a computer, cause the computer to perform the method of any of claims 1 to 8.

12. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 8.