Multimedia information delivery method and device, computer equipment and storage medium

Abstract

The invention relates to a multimedia information delivery method and device, computer equipment, a storage medium and a computer program product. The method can be applied to an application scene in which a vehicle-mounted terminal, a cloud server or other equipment interacts with a multimedia information platform. The method comprises the following steps: acquiring an application operation record, a resource interaction record, a personalized push tag, an application operation time period and activeness grouping information in a first delivery channel of each object in an application program; determining the starting probability of each object based on the application operation record; determining the interaction probability of each object based on the resource interaction record; determining a target object of a second delivery channel from the objects based on the application operation time period and the interaction probability; determining a resource delivery coefficient based on the personalized push tag, the activeness grouping information and the starting probability of the target object; and based on the resource delivery coefficient, performing multimedia information delivery of the second delivery channel to the target object. By adopting the method, the release efficiency of the multimedia information can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, apparatus, computer equipment, storage medium and computer program product for delivering multimedia information. Background Art

[0002] With the development of computer technology and Internet technology, more and more multimedia information needs to be delivered to help users quickly filter out product information that meets their needs from multimedia information. For example, when multimedia merchants deliver multimedia information, they usually need to conduct delivery experiments in different multimedia accounts in advance to determine the delivery method and delivery plan.

[0003] However, in the current multimedia information delivery method, when multimedia merchants deliver multimedia information, in order to improve the return on investment (ROI) of multimedia information delivery, they usually screen the audience based on the objects with resource interaction behaviors in the historical records and deliver multimedia information in a targeted manner, which leads to low efficiency of multimedia information delivery. Summary of the invention

[0004] Based on this, it is necessary to provide a multimedia information delivery method, device, computer equipment, computer readable storage medium and computer program product to address the above technical problems, which can effectively improve the efficiency of multimedia information delivery and at the same time improve the effect of multimedia information delivery.

[0005] In the first aspect, the present application provides a method for delivering multimedia information. The method includes: obtaining application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information in a first delivery channel for each object in an application; determining the launch probability of each object for the application based on the application operation records; determining the interaction probability of each object for resource interaction based on the resource interaction records; determining the target object of a second delivery channel from each object based on the application operation time period and the interaction probability; determining the resource delivery coefficient for multimedia information delivery based on the personalized push tags, activity grouping information, and launch probability of the target object; and delivering multimedia information of the second delivery channel to the target object based on the resource delivery coefficient.

[0006] In the second aspect, the present application also provides a multimedia information delivery device. The device includes: an acquisition module, which is used to acquire application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information in the first delivery channel for each object in the application; a determination module, which is used to determine the launch probability of each object for the application based on the application operation records; determine the interaction probability of each object for resource interaction based on the resource interaction records; determine the target object of the second delivery channel from each object based on the application operation time period and the interaction probability; determine the resource delivery coefficient for multimedia information delivery based on the personalized push tags, activity grouping information, and launch probability of the target object; and a delivery module, which delivers multimedia information of the second delivery channel to the target object based on the resource delivery coefficient.

[0007] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method in each embodiment of the present application are implemented.

[0008] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method in each embodiment of the present application are implemented.

[0009] In a fifth aspect, the present application further provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the method in each embodiment of the present application are implemented.

[0010] The above-mentioned multimedia information delivery method, device, computer equipment, storage medium and computer program product obtain the application operation records, resource interaction records, personalized push tags, application operation time period and activity grouping information of each object in the application in the first delivery channel; determine the startup probability of each object for the application based on the application operation records; determine the interaction probability of each object for resource interaction based on the resource interaction records; determine the target object of the second delivery channel from each object based on the application operation time period and the interaction probability; determine the resource delivery coefficient of the multimedia information delivery based on the personalized push tags, activity grouping information and startup probability of the target object; based on the resource delivery coefficient, deliver the multimedia information of the second delivery channel to the target object. Since the probability of each object launching an application is determined based on the application operation records of each object in the application, and the probability of each object interacting with resources is determined based on the resource interaction records of each object in the application, the target object of the second delivery channel determined based on the application operation time period of each object in the application and the interaction probability of each object interacting with resources has a higher delivery value, and the resource delivery coefficient of multimedia information delivery determined based on the personalized push tag, activity grouping information and launch probability of the target object is also more accurate, realizing more efficient population value screening and multimedia information delivery bidding, while effectively improving the delivery efficiency of multimedia information, and effectively improving the delivery effect of multimedia information. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A diagram of an application environment of a method for delivering multimedia information in an embodiment;

[0012] Figure 2 A schematic diagram of a flow chart of a method for delivering multimedia information in one embodiment;

[0013] Figure 3 A schematic diagram of group division for grouping click activity of different users in one embodiment;

[0014] Figure 4 A schematic diagram of a process in which an advertising platform conducts advertising bidding and completes an order in one embodiment;

[0015] Figure 5 A schematic diagram of a network structure of a classification model in one embodiment;

[0016] Figure 6 A schematic diagram of a multi-dimensional user tag overall application framework in one embodiment;

[0017] Figure 7 A flowchart of a step of determining a resource delivery coefficient for multimedia information delivery based on a personalized push tag, activity grouping information, and activation probability of a target object in one embodiment;

[0018] Figure 8 A schematic diagram of an algorithm architecture of an advertisement delivery method linked to self-starting behavior in one embodiment;

[0019] Fig. 9 is a structural block diagram of a multimedia information delivery device in one embodiment;

[0020] Fig.10 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0022] Cloud technology refers to a hosting technology that unifies hardware, software, network and other resources within a wide area network or local area network to achieve data computing, storage, processing and sharing.

[0023] Cloud technology is a general term for network technology, information technology, integration technology, management platform technology, application technology, etc. based on the cloud computing business model. It can form a resource pool, which is used on demand and is flexible and convenient. Cloud computing technology will become an important support. The services of technical network systems require a large amount of computing and storage resources, such as video websites, image websites and more portal websites. With the rapid development and application of the Internet industry, in the future, each item may have its own identification mark, which needs to be transmitted to the system for logical processing. Data of different levels will be processed separately. All kinds of industry data need strong system backing support, which can only be achieved through cloud computing.

[0024] Cloud storage is a new concept that extends and develops from the concept of cloud computing. A distributed cloud storage system (hereinafter referred to as storage system) refers to a storage system that uses cluster applications, grid technology, and distributed storage file systems to bring together a large number of different types of storage devices (storage devices are also called storage nodes) in the network through application software or application interfaces to work together and provide external data storage and business access functions.

[0025] With the research and advancement of artificial intelligence technology, artificial intelligence technology has been studied and applied in many fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, drones, robots, smart medical care, smart customer service, Internet of Vehicles, automatic driving, smart transportation, etc. I believe that with the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0026] It should be noted that in the following description, the terms "first, second and third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first, second and third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

[0027] The multimedia information delivery method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set separately, integrated on the server 104, or placed on the cloud or other network servers. The device of the multimedia manufacturer can be the terminal 102, and the multimedia information platform can be the server 104, that is, the device of the multimedia manufacturer, namely the terminal 102, can interact with the multimedia information platform, namely the server 104, and the terminal 102 can obtain the application operation records, resource interaction records, personalized push tags, application operation time periods and activity grouping information of each object in the application program, and determine the startup probability of each object for the application program based on the application operation record; the terminal 102 determines the interaction probability of each object for resource interaction based on the resource interaction record; the terminal 102 determines the target object of the second delivery channel from each object based on the application operation period and the interaction probability; the terminal 102 determines the resource delivery coefficient of the multimedia information delivery based on the personalized push tag, activity grouping information and startup probability of the target object; the terminal 102 can directly deliver the multimedia information of the second delivery channel to the target object based on the resource delivery coefficient. Alternatively, the terminal 102 may send the obtained object information and resource delivery coefficient of the target object to the server 104, so that the multimedia information platform compares the resource delivery coefficient with other resource delivery coefficients, and determines the result data based on the comparison result. When the result data meets the preset conditions, the server 104 may deliver multimedia information of the second delivery channel to the target object based on the resource delivery coefficient. Alternatively, the terminal 102 may send the obtained object information and resource delivery coefficient of the target object to the server 104, so that the multimedia information platform compares the resource delivery coefficient with other resource delivery coefficients, and determines the result data based on the comparison result. The terminal 102 may receive the result data returned by the server 104. When the result data meets the preset conditions, the terminal 102 delivers multimedia information of the second delivery channel to the target object based on the resource delivery coefficient.

[0028] The terminal 102 may be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart TV, a smart watch, an IoT device, and a portable wearable device. The IoT device may be a smart vehicle device, etc. The portable wearable device may be a smart watch, a smart bracelet, a head-mounted device, etc.

[0029] The server 104 may be an independent physical server or a service node in a blockchain system. A peer-to-peer network is formed between the service nodes in the blockchain system. The peer-to-peer protocol is an application layer protocol running on top of the Transmission Control Protocol (TCP).

[0030] In addition, server 104 can also be a server cluster composed of multiple physical servers, and can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN, Content Delivery NetworkCDN), as well as big data and artificial intelligence platforms.

[0031] The terminal 102 and the server 104 may be connected via Bluetooth, USB (Universal Serial Bus) or a network or other communication connection methods, which are not limited in this application.

[0032] In one embodiment, Figure 2 As shown, a method for delivering multimedia information is provided. The method can be executed by a server or a terminal alone, or by a server and a terminal together. Figure 1 The terminal in is used as an example to illustrate, including the following steps:

[0033] Step 202: Obtain application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in the application program in the first delivery channel.

[0034] Among them, the application in this application may refer to a specific type of application. For example, the application in this application may be a video application. For another example, the application in this application may be an application with resource interaction functions, such as a paid video application.

[0035] Application operation records refer to records of various types of operations performed on an application using the application's objects. For example, the application operation records in this application may include operation records of launching an application triggered by using objects, records of various interactive operations triggered in the application interface, such as click operations and drag operations triggered in the application interface, etc.

[0036] It can be understood that the application operation records in the present application include but are not limited to the above-mentioned operation records of launching the application and the records of various interactive operations triggered in the application interface.

[0037] Resource interaction records refer to records of various types of resource interactions performed by various objects of an application program on an application program. Resource interaction records may include automatically generated resource interaction records and resource interaction records actively triggered by various objects. For example, resource interaction records in this application may include active payment records, automatic renewal records, and overall payment records generated by various objects in the target application program. Among them, the overall payment record may be determined based on the active payment record and the automatic renewal record.

[0038] The personalized push tag refers to a tag of the personalized push function set by each object using the application, which is used to reflect the status of the personalized push function of each object using the application. For example, the personalized push tag in this application may include a first value and a second value. When the personalized push tag of the object using the application is the first value, it indicates that the status of the personalized push function of the object is on, indicating that the object is willing to receive various multimedia information pushed by the application through the first delivery channel; when the personalized push tag of the object using the application is the second value, it indicates that the status of the personalized push function of the object is off, indicating that the object is unwilling to receive various multimedia information pushed by the application through the first delivery channel.

[0039] The application operation period refers to the difference between the application operation time in the application operation record of each object and the target time. For example, the application operation period in this application can refer to the time difference between the operation time of the last application startup triggered by the object and the current time. The time difference is used to reflect the number of days from the last active time of the object using the application to the current time.

[0040] The first delivery channel refers to a specific delivery channel. For example, the first delivery channel in the present application may refer to a free delivery channel, and the free delivery channel may be a free delivery channel that comes with the application.

[0041] Activity group information refers to group category information divided according to the activity of each object using the application program, that is, the activity group information in this application may include group categories divided based on different activity levels, such as Figure 3 As shown in FIG. 1 , it is a schematic diagram of group division of click activity groups of different users. The activity group information in this application can be as follows: Figure 3 The classification categories shown in include six types of activity groups: non-received group, non-clicked group, random click group, low-active group, medium-active group and high-active group.

[0042] The activity grouping information in the first delivery channel refers to group category information divided based on the activity of each object in using the application in the first delivery channel.

[0043] Specifically, the devices used by different multimedia merchants can interact with the multimedia information platform. When the devices used by the multimedia merchants receive traffic request information forwarded by the multimedia information platform, the devices used by the multimedia merchants are triggered to execute the multimedia information delivery method provided in this application, that is, the devices used by the multimedia merchants can obtain application operation records, resource interaction records, personalized push tags, application operation time periods and activity grouping information of each object in the target application in the first delivery channel.

[0044] For example, a multimedia merchant is an advertiser, a multimedia information platform is an advertising platform, and a device used by an advertiser is a terminal. When the device used by advertiser A, i.e., terminal A, receives the traffic request information forwarded by the advertising platform, terminal A is triggered to respond to the traffic request information and obtain the historical operation records of each object in video application A for launching video application A, payment records, personalized push tags, application operation time period, and activity grouping information in the first delivery channel. For example, terminal A responds to the traffic request information and obtains the historical operation records of launching video application A, the historical overall payment records, the personalized push tag as the first value, the application operation time period as 9 days, and the activity grouping information in the first delivery channel as the medium active group of object A in video application A.

[0045] Step 204: Determine the probability of each object starting the application program based on the application operation record.

[0046] Among them, the startup probability refers to the probability of an object starting an application. For example, the startup probability in this application can be the probability of each object actively opening the application on the next day predicted by the neural network model.

[0047] Specifically, after the terminal obtains the application operation records, resource interaction records, personalized push tags, application operation time periods and activity grouping information of each object in the application, the terminal can determine the launch probability of each object for the application based on the application operation records. For example, the terminal can obtain the application operation records of each object (operation records for launching the application) from the background of the target application, and process the obtained application operation records (operation records for launching the application) of each object through a pre-trained machine learning model or a deep learning model to obtain the launch probability of each object for the application. That is, in this application, the obtained application operation records (operation records for launching the application) of each object can be processed through a pre-trained machine learning model or a deep learning model, and the output parameter is the launch probability of each object for the application.

[0048] For example, a multimedia merchant is an advertiser, a multimedia information platform is an advertising platform, and a device used by an advertiser is a terminal. Assuming that terminal A obtains the application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in video application A based on the traffic request information forwarded by the advertising platform, terminal A can process the obtained application operation records (operation records of launching applications) of each object through a pre-trained machine learning model or a deep learning model, and obtain the launch probability of each object for the application. For example, assuming that terminal A obtains the historical launch operation records of user A in video application A, the historical overall payment records, the personalized push tag is the first value, the application operation period is 9 days, and the activity grouping information in the first delivery channel is a medium active group based on the traffic request information forwarded by the advertising platform, terminal A can process the historical launch operation records of user A in video application A through a pre-trained machine learning model or a deep learning model, and obtain the launch probability of user A for video application A.

[0049] Step 206: Determine the probability of each object interacting with the resource based on the resource interaction record.

[0050] The interaction probability refers to the probability of each object interacting with resources in the application. The interaction probability in this application may include different types of interaction probabilities. For example, the interaction probability in this application may be a payment probability, and the payment probability may include an automatic renewal probability, an active payment probability, and an overall payment probability. That is, the interaction probability in this application may be the overall payment probability of each object in the application.

[0051] Specifically, after the terminal determines the probability of each object starting the application based on the application operation record, the terminal can determine the interaction probability of each object performing resource interaction based on the resource interaction record. For example, the terminal can obtain the historical resource interaction records (historical payment records) of each object from the background of the target application, and process the obtained historical resource interaction records (historical payment records) of each object through a pre-trained machine learning model or a deep learning model to obtain the interaction probability (overall payment probability) of each object performing resource interaction in the target application. That is, in this application, the obtained historical resource interaction records (historical payment records) of each object can be processed through a pre-trained machine learning model or a deep learning model, and the output parameter is the interaction probability (overall payment probability) of each object performing resource interaction in the target application.

[0052] For example, a multimedia merchant is an advertiser, a multimedia information platform is an advertising platform, and a device used by an advertiser is a terminal. Assuming that terminal A obtains the application operation records, resource interaction records, personalized push tags, application operation periods, and activity grouping information of each object in video application A based on the traffic request information forwarded by the advertising platform, terminal A can process the historical resource interaction records (historical payment records) of each object obtained through a pre-trained machine learning model or deep learning model to obtain the interaction probability of each object performing resource interaction in the target application. For example, assuming that terminal A obtains the historical operation records of user A starting video application A in video application A based on the traffic request information forwarded by the advertising platform, the historical payment records are record A, the personalized push tags are the first value, the application operation period is 9 days, and the activity grouping information in the first delivery channel is a medium active group, terminal A can process the historical payment records of user A, i.e., record A, through a pre-trained machine learning model or deep learning model to obtain the predicted payment probability of user A in video application A.

[0053] Step 208: Determine the target object of the second delivery channel from among the objects based on the application operation period and the interaction probability.

[0054] The second delivery channel refers to a specific delivery channel. For example, the first delivery channel in this application refers to a free delivery channel, which may be a free delivery channel provided by the application. The second delivery channel may refer to a paid delivery channel, such as a paid delivery channel may be an RTA (Real Time API) channel. It can be understood that the first delivery channel and the second delivery channel in this application are only used to distinguish delivery channels of different multimedia information, and no specific limitation is required in this application.

[0055] The target object refers to an object for the second delivery channel. For example, the target object in the present application may be an object that is screened out and is more suitable for a paid delivery channel.

[0056] Specifically, after the terminal determines the interaction probability of each object for resource interaction based on the resource interaction record, the terminal can determine the target object of the second delivery channel from each object based on the application operation period and interaction probability of each object. For example, the terminal can screen out the target object that is more suitable for the paid delivery channel from all objects using the application based on the application operation period and predicted payment probability of each object. The target object can be one or more. That is, the terminal can compare the application operation period of each object with the preset time threshold. When the application operation period of the user is less than or equal to the preset time threshold, the terminal can further compare the predicted payment probability of each user with the preset probability value to obtain a comparison result, and determine the target object of the paid delivery channel based on the comparison result.

[0057] For example, multimedia merchants are advertisers, multimedia information platforms are advertising platforms, and the devices used by advertisers are terminals. Assuming that terminal A processes the payment records of each user in video application A through a pre-trained machine learning model or deep learning model, and obtains the predicted payment probability of each user in video application A, terminal A can determine the target object of the paid delivery channel based on the application operation period and predicted payment probability of each user. For example, terminal A can compare the application operation period of each user with a preset time threshold. When the application operation period of the user is less than or equal to the preset time threshold, terminal A can compare the predicted payment probability of each user with the preset probability value to obtain a comparison result, and determine the target object of the second delivery channel based on the comparison result. Assuming that the application operation period of user A is less than the preset time threshold, and the interaction probability (predicted payment probability) of user A is greater than the preset probability value, terminal A can use user A as the target object of the paid delivery channel.

[0058] Step 210: Determine a resource delivery coefficient for multimedia information delivery based on the personalized push tag, activity grouping information, and activation probability of the target object.

[0059] Among them, the resource delivery coefficient refers to the resource delivery coefficient used for multimedia information delivery, and the resource delivery coefficient is used to reflect the amount of resources used for multimedia information delivery. For example, when the resource delivery coefficient is larger, it means that more resources are used for multimedia information delivery; when the resource delivery coefficient is smaller, it means that fewer resources are used for multimedia information delivery. It can be understood that the resource delivery coefficient in this application can be a bid coefficient, and the bid coefficient is used to reflect the price for multimedia information delivery, that is, when the bid coefficient is larger, it means that the price for multimedia information delivery is higher; when the bid coefficient is smaller, it means that the price for multimedia information delivery is lower.

[0060] Step 212: Based on the resource delivery coefficient, multimedia information is delivered to the target object through the second delivery channel.

[0061] Specifically, after the terminal determines the target object of the second delivery channel from each object based on the application operation period and the interaction probability, the terminal can determine the resource delivery coefficient of the multimedia information delivery based on the personalized push tag, activity grouping information and activation probability of the target object, and directly deliver the multimedia information of the second delivery channel to the target object based on the determined resource delivery coefficient. Among them, there are many ways to determine the resource delivery coefficient of the multimedia information delivery based on the personalized push tag, activity grouping information and activation probability of the target object in this application, which are not specifically limited here.

[0062] It can be understood that in some cases, the terminal may also interact with the multimedia delivery platform, that is, after the terminal determines the target object of the second delivery channel from various objects based on the application operation period and the interaction probability, the terminal may determine the resource delivery coefficient A for the multimedia information delivery based on the personalized push tag, activity grouping information and startup probability of the target object; further, the terminal may package the object information and the resource delivery coefficient A of the target object determined by the terminal together and send them to the multimedia delivery platform, so that the multimedia delivery platform compares the resource delivery coefficient A for the target object sent by the terminal with the resource delivery coefficient B sent by other terminals (multimedia merchants), and judges based on the comparison result whether to finally deliver the multimedia information of the second delivery channel to the target object based on the resource delivery coefficient A sent by the terminal.

[0063] For example, a multimedia merchant is an advertiser, a multimedia information platform is an advertising platform, and a device used by the advertiser is a terminal. Assuming that after terminal A determines that the target object of the paid delivery channel is user A, terminal A can determine that the bid coefficient for multimedia information delivery is A1 based on the personalized push tag, activity grouping information, and predicted startup probability of user A in video application A, and then terminal A can directly deliver multimedia information of the paid delivery channel to user A based on the bid coefficient A1. Alternatively, after terminal A determines that the bid coefficient for multimedia information delivery is A1 based on the personalized push tag, activity grouping information, and predicted startup probability of user A in video application A, terminal A can package the bid coefficient A1 and the information of the target object, i.e. user A, together and send them to the advertising platform, so that the advertising platform compares the bid coefficient A1 for the target object sent by the terminal A with the bid coefficient B1 sent by other terminals (multimedia merchants), and determines whether to deliver multimedia information of the paid delivery channel to the target object, i.e. user A, based on the comparison result. Assuming that A1 is greater than B1, the advertising platform determines that the bid coefficient A1 sent by terminal A has successfully bid based on the comparison result. The advertising platform will eventually deliver multimedia information to the target object, i.e. user A, through the paid delivery channel based on the bid coefficient A1 sent by terminal A, and return the corresponding delivery effect data to terminal A.

[0064] In this embodiment, the application operation records, resource interaction records, personalized push tags, application operation time periods and activity grouping information in the first delivery channel of each object in the application are obtained; the startup probability of each object for the application is determined based on the application operation records; the interaction probability of each object for resource interaction is determined based on the resource interaction records; the target object of the second delivery channel is determined from each object based on the application operation time period and the interaction probability; the resource delivery coefficient for multimedia information delivery is determined based on the personalized push tags, activity grouping information and startup probability of the target object; and based on the resource delivery coefficient, multimedia information of the second delivery channel is delivered to the target object. Since the probability of each object launching an application is determined based on the application operation records of each object in the application, and the probability of each object interacting with resources is determined based on the resource interaction records of each object in the application, the target object of the second delivery channel determined based on the application operation time period of each object in the application and the interaction probability of each object interacting with resources has a higher delivery value, and the resource delivery coefficient of multimedia information delivery determined based on the personalized push tag, activity grouping information and launch probability of the target object is also more accurate, realizing more efficient population value screening and multimedia information delivery bidding, while effectively improving the delivery efficiency of multimedia information, and effectively improving the delivery effect of multimedia information.

[0065] In one embodiment, the method further comprises:

[0066] Sending the object information and resource placement coefficient of the target object to the multimedia information platform, so that the multimedia information platform compares the resource placement coefficient with other resource placement coefficients and determines result data based on the comparison result;

[0067] Receive result data returned by the multimedia information platform;

[0068] The delivering the multimedia information of the second delivery channel to the target object based on the resource delivery coefficient includes:

[0069] When the result data meets the preset conditions, multimedia information is delivered to the target object through the second delivery channel based on the resource delivery coefficient.

[0070] Among them, other resource delivery coefficients refer to resource delivery coefficients sent by other terminals (other multimedia manufacturers). For example, the other resource delivery coefficients in this application can be resource delivery coefficients sent by devices of other multimedia manufacturers to the multimedia information platform. It can be understood that the other resource delivery coefficients and resource delivery coefficients in this application are only used to distinguish between resource delivery coefficients determined by different multimedia manufacturers. The other resource delivery coefficients and resource delivery coefficients can be the same or different. For example, the other resource delivery coefficient can be A1, and the resource delivery coefficient determined by this terminal can also be A1.

[0071] The comparison result refers to the comparison result between other resource delivery coefficients and the resource delivery coefficients determined by this terminal. For example, the comparison result in this application may include a sorting sequence obtained by sorting the resource delivery coefficients sent by various multimedia manufacturers in order from large to small. The comparison result in this application may also include the size relationship between other resource delivery coefficients and the resource delivery coefficients.

[0072] Result data refers to data used to reflect the comparison results. For example, the result data in this application may include the size relationship between other resource allocation coefficients and resource allocation coefficients.

[0073] The preset condition refers to a pre-set condition for determining whether to deliver multimedia information to the second delivery channel. For example, the preset condition in the present application is: a condition for successful bidding. The condition for successful bidding may be that a resource delivery coefficient is greater than other resource delivery coefficients.

[0074] Specifically, take multimedia merchants as advertisers, multimedia information platforms as advertising platforms, and devices used by advertisers as terminals as examples for explanation. Assuming that terminal A responds to the traffic request information forwarded by the advertising platform, and obtains the application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in the video application A in the first delivery channel, terminal A can process the obtained application operation records (operation records for launching applications) and historical resource interaction records (historical payment records) of each object through a pre-trained machine learning model or deep learning model, and obtain the probability of launching the application for each object and the probability of each object interacting with resources in the target application; further, terminal A can determine the target object of the paid delivery channel based on the application operation time period and predicted payment probability of each user, and determine the resource delivery coefficient of multimedia information delivery as A1 based on the personalized push tags, activity grouping information, and launch probability of the target object; further, terminal A can process the target object The object information and resource delivery coefficient A1 of the target object are sent to the multimedia information platform, so that the multimedia information platform compares the resource delivery coefficient A1 with other resource delivery coefficients (B1, C1), and determines the result data as (A1>B1>C1) based on the comparison result (A1>B1>C1); the multimedia information platform can return the result data (A1>B1>C1) to terminal A, that is, when terminal A receives the result data (A1>B1>C1) returned by the multimedia information platform, terminal A can further determine whether the result data (A1>B1>C1) returned by the multimedia information platform meets the preset conditions. Assuming that the preset condition is that the resource delivery coefficient is greater than other resource delivery coefficients, since (A1>B1>C1), when the result data (A1>B1>C1) meets the preset conditions, terminal A can directly deliver multimedia information to the target object through the paid delivery channel based on the resource delivery coefficient A1. This enables more efficient population value screening and advertising bidding, that is, a one-stop buying decision, which effectively improves the efficiency of multimedia information delivery while effectively improving the effectiveness of multimedia information delivery.

[0075] In one embodiment, the method further comprises:

[0076] Receiving traffic request information sent by the multimedia information platform; the traffic request information is sent by the multimedia information platform to the client supporting the second delivery channel;

[0077] The obtaining of application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in the application program in the first delivery channel includes:

[0078] In response to the traffic request information, obtain application operation records and resource interaction records of each object in the application, and obtain personalized push tags of each object and activity grouping information in the first delivery channel;

[0079] Determine the application operation period based on the application operation record.

[0080] The client refers to the equipment of various multimedia manufacturers. For example, the client in this application can be the terminal of various multimedia manufacturers.

[0081] Specifically, Figure 4 As shown in FIG. 1 , it is a schematic diagram of the process of an advertising platform conducting an advertising bidding and completing an order. The multimedia merchant is used as the advertiser, the multimedia information platform is used as the advertising platform, and the device used by the advertiser is used as the terminal. Figure 4 As shown in , when the advertising platform obtains a traffic access request, it forwards the traffic access request information (including user equipment, ad space, etc.) to advertisers that support RTA (RTA requests external advertisers), and the advertisers decide whether to participate in the bidding of the platform, that is, whether to display ads under the current traffic, and what the bidding price is. For each advertiser, each successful bid requires the payment of advertising fees. The higher the ROI (return on investment), the more willing the advertiser is to participate in the bidding. After the advertising platform forwards the traffic access request information (including user devices, advertising positions, etc.) to each advertiser supporting RTA, for example, when the terminal A of the advertising manufacturer A receives the traffic access request information forwarded by the advertising platform, the terminal A responds to the traffic access request information forwarded by the advertising platform, obtains the application operation records, resource interaction records, personalized push tags and activity grouping information of each object in the target application A, and further, the terminal A can determine the application operation period according to the application operation records of each object in the target application A obtained, for example, the terminal A can obtain the target time from the application operation records of each object, and determine the difference between the target time and the current time, and use the difference as the application operation period of each object. Thus, by constructing multi-dimensional object features, that is, constructing the user's self-start probability, click activity, payment probability and other features based on the user's historical behavior through statistical methods, deep learning models and other methods, and formulating application rules based on multi-dimensional object features, it is possible to implement more accurate and rapid population value screening and advertising bidding, that is, to achieve a one-stop purchase decision, while effectively improving the delivery efficiency of multimedia information, and effectively improving the delivery effect of multimedia information.

[0082] In one embodiment, the step of obtaining personalized push tags of each object and activity grouping information in the first delivery channel includes:

[0083] Get personalized push tags for each object in the application;

[0084] Obtain the number of pushed information and clicked information for each object in the first delivery channel;

[0085] Based on the number of pushed information, the number of clicked information and the time information, the activity grouping information of each object in the first delivery channel is determined.

[0086] Among them, the number of push information refers to the number of push information received by each object in the first delivery channel. For example, the number of push information in this application can be the number of push information received by each object in the free delivery channel of the application within the past month.

[0087] The number of click information refers to the number of click information actively triggered by each object in the first delivery channel. For example, the number of click information in this application can be the number of click information actively triggered by each object in the free delivery channel of the application within the past month.

[0088] Time information refers to a preset time range. For example, the time information in this application may include different time period information such as a first preset time period, a second preset time period, and a third preset time period.

[0089] Specifically, a multimedia merchant is an advertiser, a multimedia information platform is an advertising platform, and a device used by an advertiser is a terminal. After receiving the traffic request information sent by the multimedia information platform, the terminal A responds to the traffic request information forwarded by the advertising platform, and the step of obtaining the personalized push tags of each object in the video application A and the activity grouping information in the first delivery channel includes: the terminal A can directly obtain the personalized push tags of each object in the video application A, and obtain the number of push information and click information of each object in the free delivery channel in the past month. The terminal A can determine the activity grouping information of each object in the free delivery channel based on the number of push information, the number of click information and the preset information of each time period. As a result, by constructing multi-dimensional object features, that is, constructing the user's self-start probability, click activity, payment probability and other features based on the user's historical behavior through statistical methods, deep learning models and other methods, and formulating application rules based on multi-dimensional object features, it is possible to implement more accurate and rapid population value screening and advertising bidding, that is, to achieve a one-stop purchase decision, while effectively improving the delivery efficiency of multimedia information, and effectively improving the delivery effect of multimedia information.

[0090] In one embodiment, the step of determining the activity grouping information of each object in the first delivery channel based on the number of pushed information, the number of clicked information and time information includes:

[0091] When the time information is a first preset time period and the number of click information is greater than a first preset value and not greater than a second preset value, it is determined that the object is in a first active group in the first delivery channel;

[0092] When the time information is a second preset time period and the number of click information is greater than the first preset value and not greater than the second preset value, it is determined that the object is in the second active group in the first delivery channel;

[0093] When the time information is a second preset time period and the number of click information is greater than a second preset value, it is determined that the object is in the third active group in the first delivery channel;

[0094] Among them, the activity levels of the first active group, the second active group, and the third active group increase in sequence.

[0095] The first preset time period and the second preset time period are only used to distinguish different preset time periods. For example, the first preset time period in this application can be as follows: Figure 3 At least one month of the most recent n months shown in ; the second preset time period in this application can be as follows Figure 3 Each of the last n months shown in .

[0096] The first preset value and the second preset value are only used to distinguish different preset values. For example, Figure 3 As shown in , the first preset value in the present application may be a, and the second preset value may be b.

[0097] The first active group, the second active group and the third active group in this application are only used to distinguish different activity groups, for example Figure 3 As shown in , the first active group in the present application may be a low-activity group, the second active group may be a medium-activity group, and the third active group may be a high-activity group.

[0098] Specifically, a multimedia merchant is an advertiser, a multimedia information platform is an advertising platform, and a device used by the advertiser is a terminal. Terminal A responds to the traffic request information forwarded by the advertising platform and obtains the number of push information and click information of each object in the free delivery channel in the past month. Terminal A can determine the activity grouping information of each object in the free delivery channel based on the number of push information, the number of click information and the preset time period information, for example, Figure 3As shown, when the number of click information of object A in at least one of the last n months is greater than a and not greater than b, terminal A determines that object A is in the low-activity group in the first delivery channel; when the number of click information of object B in each of the last n months is greater than a and not greater than b, terminal A determines that object B is in the medium-activity group in the first delivery channel; when the number of click information of object C in each of the last n months is greater than b, terminal A determines that object C is in the high-activity group in the first delivery channel. Therefore, by constructing multi-dimensional object features, that is, constructing the user's self-start probability, click activity, payment probability and other features based on the user's historical behavior through statistical methods, deep learning models and other methods, and formulating application rules based on multi-dimensional object features, it is possible to implement more accurate and rapid population value screening and advertising bidding, that is, to achieve a one-stop purchase decision, while effectively improving the delivery efficiency of multimedia information, and effectively improving the delivery effect of multimedia information.

[0099] In one embodiment, the step of determining the activity grouping information of each object in the first delivery channel based on the number of pushed information, the number of clicked information and time information includes:

[0100] When the time information is a third preset time period and the number of pushed information pieces is a third preset value, it is determined that the object is in a fourth active group in the first delivery channel;

[0101] When the time information is a third preset time period, the number of pushed information is not a third preset value, and the number of clicked information is a third preset value, it is determined that the object is in the fourth active group in the first delivery channel;

[0102] When the time information is a second preset time period and the number of click information is not greater than a second preset value, it is determined that the object is in the fourth active group in the first delivery channel; the activity levels of the fourth active group, the first active group, the second active group, and the third active group increase sequentially.

[0103] The first preset time period, the second preset time period, and the third preset time period are only used to distinguish different preset time periods. For example, the third preset time period in this application can be as follows: Figure 3 The most recent month shown in .

[0104] The first preset value, the second preset value, and the third preset value are only used to distinguish different preset values. For example, Figure 3 As shown in , the third preset value in the present application may be 0.

[0105] The first active group, the second active group, the third active group, and the fourth active group in this application are only used to distinguish different activity groups, for example, Figure 3As shown in , the fourth active group in the present application may be an inactive group, and the inactive group may be further divided into: a not received, a not clicked, and a random clicked group.

[0106] Specifically, a multimedia merchant is an advertiser, a multimedia information platform is an advertising platform, and a device used by the advertiser is a terminal. Terminal A responds to the traffic request information forwarded by the advertising platform and obtains the number of push information and click information of each object in the free delivery channel in the past month. Terminal A can determine the activity grouping information of each object in the free delivery channel based on the number of push information, the number of click information and the preset time period information, for example, Figure 3 As shown, when the number of push messages of object A in the last month is 0, terminal A determines that object A is in the non-receiving group in the inactive group in the first delivery channel; when the number of push messages of object B in the last month is not 0 and the number of click messages is 0, terminal A determines that object B is in the non-clicking group in the inactive group in the first delivery channel; when the number of click messages of object C in each of the last n months is not greater than a, terminal A determines that object C is in the random clicking group in the inactive group in the first delivery channel. Therefore, by constructing multi-dimensional object features, that is, constructing the user's self-starting probability, click activity, payment probability and other features based on the user's historical behavior through statistical methods, deep learning models and other methods, and formulating application rules based on multi-dimensional object features, it is possible to implement more accurate and rapid population value screening and advertising bidding, that is, to achieve a one-stop buying volume decision, which effectively improves the delivery efficiency of multimedia information while effectively improving the delivery effect of multimedia information.

[0107] In one embodiment, the step of obtaining personalized push tags of each object and activity grouping information in the first delivery channel includes:

[0108] Obtain activity grouping information of each object in the first delivery channel;

[0109] Detect the on / off status of the personalized push function for each object in the application;

[0110] Determine the personalized push label for each object based on the open and closed status.

[0111] Among them, the on-off state refers to whether the personalized push function is in an on state or an off state. For example, when the on-off state of the personalized push function is on, the personalized push tag of the object is determined to be 0; when the on-off state of the personalized push function is off, the personalized push tag of the object is determined to be 1.

[0112] Specifically, take the multimedia merchant as the advertiser, the multimedia information platform as the advertising platform, and the device used by the advertiser as the terminal as an example for explanation. After the terminal A obtains the activity grouping information of each object in the first delivery channel in response to the traffic request information forwarded by the advertising platform, the terminal A can detect the on / off state of the personalized push function of each object in the target application, and determine the personalized push label of each object according to the on / off state. For example, when it is detected that the on / off state of the personalized push function of object A in the target application is on, the terminal A can determine that the personalized push label of object A is 0; when it is detected that the on / off state of the personalized push function of object A in the target application is off, the terminal A can determine that the personalized push label of object A is 1. Whether the user turns on the switch of the personalized push function in the application belongs to a fact label. If the switch is on, the user can be reached through the free channel, otherwise it cannot be reached. In the embodiment of the present application, the personalized push label of the target object is used as one of the influencing factors of the bid coefficient affecting the multimedia information delivery, which can effectively improve the accuracy of the bid coefficient, thereby effectively improving the delivery efficiency of multimedia information, and at the same time improving the delivery effect of multimedia information.

[0113] In one embodiment, the step of determining the application operation period according to the application operation record includes:

[0114] Get the target time from the application operation record;

[0115] Determine the difference between the target time and the current time;

[0116] The difference is taken as the application operation period of each object.

[0117] The target time refers to a specific time, for example, the target time in this application may be the time when each object last operated on the application. The application operation time may be the time when each object starts the application program.

[0118] Specifically, take the multimedia merchant as the advertiser, the multimedia information platform as the advertising platform, and the device used by the advertiser as the terminal as an example. After terminal A receives the traffic request information sent by the multimedia information platform, terminal A responds to the traffic request information forwarded by the advertising platform, obtains the application operation records and resource interaction records of each object in the target application, and obtains the personalized push tags of each object and the activity grouping information in the first delivery channel. Terminal A can obtain the time of the most recent (last) application startup from the application operation record (operation record of launching the application), and calculates the difference between the time and the current time, and uses the difference as the application operation period of each object. For example, terminal A can obtain the time when user A last (last) launched application A from the application operation record (operation record of launching the application) as October 18, 2023, and calculate the difference between this time and the current time October 19, 2023. Assuming that the calculation unit is days, terminal A calculates the time difference between October 19, 2023 and October 18, 2023 as 1 day, and uses the time difference 1 as the application operation period of user A. That is, in the embodiment of the present application, by counting the number of days from the last time each user launched the application to the current time, the value range is a natural number of [1, ∞], and the larger the value of the difference, the lower the user activity. Therefore, by taking the application operation period of the target object as one of the influencing factors affecting the bid coefficient of multimedia information delivery, the accuracy of the bid coefficient can be effectively improved, thereby effectively improving the delivery efficiency of multimedia information, and at the same time improving the delivery effect of multimedia information.

[0119] In one embodiment, determining the probability of each object starting the application program based on the application operation record includes:

[0120] The application operation records of each object are processed through a deep neural network model to obtain the startup probability of each object for the application.

[0121] Among them, the deep neural network model in the present application can be a pre-trained model for predicting the probability of a user actively opening an application the next day.

[0122] Specifically, after the terminal obtains the application operation records, resource interaction records, personalized push tags, application operation time periods and activity grouping information of each object in the application, the terminal can obtain the application operation records (operation records of launching the application) of each object from the background of the target application, and process the obtained application operation records (operation records of launching the application) of each object through a pre-trained deep learning model, and obtain the probability of launching the application for each object, that is, in this application, the obtained application operation records (operation records of launching the application) of each object can be processed by a pre-trained deep learning model, and the output parameter is the probability of launching the application for each object. Among them, the deep learning model used in this application is obtained by pre-training, that is, the historical active records of different users are used as training samples in advance, and the initial deep learning model is trained to obtain the trained deep learning model, wherein the method used in the training is DNN (Deep Nueral Network), and the training target is binary classification. After the model training is completed, a given user group can be scored, and the value range is [0, 1]. The higher the prediction score, the higher the probability that the user actively launches the application. Therefore, by using a pre-trained deep learning model to process the application operation records of each object (operation records of launching the application), the probability of each user launching the application can be obtained quickly and accurately, and the probability of launching the application can be used as one of the influencing factors of the bid coefficient affecting the multimedia information delivery. This can effectively improve the accuracy of the bid coefficient, and thus effectively improve the efficiency of multimedia information delivery, while also improving the effect of multimedia information delivery.

[0123] In one embodiment, the step of determining the interaction probability of each object performing resource interaction based on the resource interaction record includes:

[0124] Classify the resource interaction records of each object through each network model in the classification model to obtain a first resource interaction probability and a second resource interaction probability;

[0125] The interaction probability of each object performing resource interaction is determined according to the first resource interaction probability and the second resource interaction probability.

[0126] Among them, the first resource interaction probability and the second resource interaction probability are only used to distinguish different types of resource interaction probabilities. For example, the first resource interaction probability in the present application can be the automatic renewal probability, and the second resource interaction probability can be the active payment probability. The interaction probability of each object for resource interaction can be the overall payment probability of each object in the application, that is, the overall payment probability = automatic renewal probability + active payment probability.

[0127] Specifically, after the terminal determines the probability of each object starting the application based on the application operation record, the terminal can classify the resource interaction records (historical payment records) of each object through the network models in the classification model to obtain the first resource interaction probability (automatic renewal probability) and the second resource interaction probability (active payment probability) of each object. Furthermore, the terminal can calculate the sum of the first resource interaction probability (automatic renewal probability) and the second resource interaction probability (active payment probability) of each object, and use the sum as the interaction probability (overall payment probability) of each object for resource interaction.

[0128] For example, a multimedia merchant is an advertiser, a multimedia information platform is an advertising platform, and a device used by an advertiser is a terminal. Assuming that terminal A obtains the application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in video application A based on the traffic request information forwarded by the advertising platform, terminal A can process the obtained historical payment records of each object through the two-way network model in the pre-trained classification model, and obtain the payment probability of each object in the target application. For example, assuming that terminal A obtains the historical operation records of user A in video application A starting video application A, the historical payment records are record A, the personalized push tags are the first value, the application operation period is 9 days, and the activity grouping information in the first delivery channel is a medium active group based on the traffic request information forwarded by the advertising platform, terminal A can process the historical payment records of user A, i.e., record A, through the two-way network model in the pre-trained classification model, and obtain the overall payment probability of user A in video application A predicted by the model.

[0129] In this embodiment, by using a pre-trained classification model to process the historical payment records of each object, the predicted payment probability of each user can be obtained quickly and accurately, and the predicted payment probability can be used as one of the influencing factors of the bid coefficient affecting the multimedia information delivery. This can effectively improve the accuracy of the bid coefficient, and thus effectively improve the efficiency of multimedia information delivery, while also improving the effect of multimedia information delivery.

[0130] In one embodiment, before classifying resource interaction records of each object by each network model in the classification model to obtain the first resource interaction probability and the second resource interaction probability, the method further includes:

[0131] Acquire a training sample including a first resource interaction and a second resource interaction;

[0132] The first network model and the second network model in the initial classification model are trained based on the training samples until the sum of the first loss value of the first network model and the second loss value of the second network model meets the loss value condition, thereby obtaining a classification model.

[0133] Among them, the first resource interaction and the second resource interaction are only used to distinguish different types of resource interactions. For example, the first resource interaction in this application can be automatic renewal, and the second resource interaction can be active payment. The training sample in this application can be the user's historical payment record, and the user's historical payment record can include automatic renewal records and active payment records.

[0134] The first network model and the second network model are only used to distinguish different types of network models. For example, the first network model in the present application may be a network for identifying the active payment probability part, and the second network model may be a network for identifying the automatic payment probability part.

[0135] Specifically, Figure 5 As shown in the figure, it is a schematic diagram of the network structure of the classification model. Since member payment in the application is a renewal behavior and non-member payment is a first-time payment behavior, the two have great differences in behavior. Therefore, when training the model, they are learned as two tasks, that is, the automatic renewal task (binary classification, whether to renew) and the active payment task (binary classification, whether to pay for the first time) are learned at the same time. The automatic renewal samples only affect the network of the automatic renewal probability part, and the active payment samples only affect the network of the active payment probability part, such as Figure 5 The two probabilities shown in are added together to output the overall payment probability. That is, the terminal can obtain training samples containing the first resource interaction and the second resource interaction. For example, the training samples can be 100 user history records containing automatic renewal records, active payment records and no payment records; further, the terminal can train the first network model and the second network model in the initial classification model based on the training samples until the sum of the first loss value of the first network model and the second loss value of the second network model meets the loss value condition, and then the trained classification model can be obtained. Among them, for the first network model, active payment records can be used as positive samples and automatic renewal records can be used as negative samples; for the second network model, active payment records can be used as negative samples and automatic renewal records can be used as positive samples.

[0136] In order to further solve the problem of scarce positive samples, the loss functions of the first network model and the second network model in the classification model of this application can adopt focal loss. It is modified based on the standard cross entropy loss. This loss function can reduce the weight of easy-to-classify samples, so that the network model can focus more on difficult-to-classify samples during training. The loss function is shown in the following formula (1):

[0137] FL(p t )=-α t (1-p t ) γ log (p t ) (1)

[0138] Among them, α is used to control the weight of positive and negative samples. α takes a relatively small value to reduce the weight of negative samples (the type of samples with more samples); γ is used to control the weight of difficult and easy samples. The purpose is to reduce the weight of easy-to-separate samples, so that the network model can focus more on the learning of difficult-to-separate samples during training.

[0139] In this embodiment, by using a pre-trained classification model to process the historical payment records of each object, the predicted payment probability of each user can be obtained quickly and accurately, and the predicted payment probability can be used as one of the influencing factors of the bid coefficient affecting the multimedia information delivery. This can effectively improve the accuracy of the bid coefficient, and thus effectively improve the efficiency of multimedia information delivery, while also improving the effect of multimedia information delivery.

[0140] In one embodiment, the network model includes a first network model and a second network model; and the step of classifying resource interaction records of each object by each network model in the classification model to obtain a first resource interaction probability and a second resource interaction probability includes:

[0141] Classify and process the resource interaction records of each object through the first network model to obtain a first resource interaction probability of each object performing resource interaction;

[0142] The resource interaction records of each object are classified and processed by the second network model to obtain a second resource interaction probability of each object performing resource interaction.

[0143] Specifically, take multimedia merchants as advertisers, multimedia information platforms as advertising platforms, and devices used by advertisers as terminals as examples for explanation. Assuming that terminal A obtains application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information in the first delivery channel of each object in video application A based on the traffic request information forwarded by the advertising platform, terminal A can process the historical payment records of each object obtained through the two-way network model in the pre-trained classification model, that is, classify the resource interaction records of each object through the first network model in the pre-trained classification model to obtain the first resource interaction probability of each object for resource interaction; classify the resource interaction records of each object through the second network model in the pre-trained classification model to obtain the second resource interaction probability of each object for resource interaction. Further, terminal A can calculate the sum of the first resource interaction probability and the second resource interaction probability of each object, and use the sum as the payment probability of each object in the target application, so as to obtain the overall payment probability of each object in video application A predicted by the model.

[0144] In this embodiment, by using a pre-trained classification model to process the historical payment records of each object, the predicted payment probability of each user can be obtained quickly and accurately, and the predicted payment probability can be used as one of the influencing factors of the bid coefficient affecting the multimedia information delivery. This can effectively improve the accuracy of the bid coefficient, and thus effectively improve the efficiency of multimedia information delivery, while also improving the effect of multimedia information delivery.

[0145] In one embodiment, the step of determining the target object of the second delivery channel from various objects based on the application operation period and the interaction probability includes:

[0146] When the application operation period is less than or equal to the preset time threshold and the interaction probability is greater than the preset probability value, the object whose application operation period is less than or equal to the preset time threshold and the interaction probability is greater than the preset probability value is taken as the target object of the second delivery channel.

[0147] The preset time threshold refers to a preset threshold for filtering application operation time periods. For example, the preset time threshold in the present application may be 90 (unit: days).

[0148] The preset probability value refers to a preset threshold value for screening interaction probabilities. For example, the preset probability value in the present application may be 0.5.

[0149] Specifically, take multimedia merchants as advertisers, multimedia information platforms as advertising platforms, and devices used by advertisers as terminals as examples for explanation. Assuming that terminal A responds to the traffic request information forwarded by the advertising platform, and obtains the application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in the video application A in the first delivery channel, terminal A can process the obtained application operation records (operation records for launching applications) and historical resource interaction records (historical payment records) of each object through a pre-trained machine learning model or deep learning model, and obtain the probability of launching the application for each object and the probability of interacting with resources for each object in the target application; further, if Figure 6 As shown, it is a schematic diagram of the overall application framework of multi-dimensional user tags. Terminal A can compare the application operation period of each object (the number of days from the last active time to today) with the preset time threshold of 90 days. When the application operation period of the object (the number of days from the last active time to today) is greater than the preset time threshold of 90 days, it is determined that the object does not belong to the target object of the second delivery channel (i.e., not released); when the application operation period of the object (the number of days from the last active time to today) is less than or equal to the preset time threshold of 90 days, terminal A can compare the interaction probability (i.e., payment probability) of each object with a preset probability value of 0.5. When the interaction probability (i.e., payment probability) of the object is less than or equal to the preset probability value of 0.5, it is determined that the object does not belong to the target object of the second delivery channel (i.e., not released); when the interaction probability (i.e., payment probability) of the object is greater than the preset probability value of 0.5, it is determined that the object is the target object of the second delivery channel. That is, terminal A takes the object whose application operation period is less than or equal to the preset time threshold of 90 days and whose interaction probability is greater than the preset probability value of 0.5 as the target object of the second delivery channel. As a result, the target population of the paid delivery channel can be determined more accurately based on the application operation period and interaction probability of each object, which not only achieves high-precision population screening, but also realizes the scale of multimedia information delivery. While achieving the ideal delivery effect, it also effectively improves the efficiency of multimedia information delivery.

[0150] In one embodiment, Figure 7 As shown, the step of determining the resource delivery coefficient for multimedia information delivery based on the personalized push tag, activity grouping information and activation probability of the target object includes:

[0151] Step 702: When the personalized push tag of the target object is a third preset value, and the activity grouping information and the activation probability are less than or equal to a preset condition threshold, a first resource delivery coefficient for multimedia information delivery is obtained.

[0152] Step 704, when the personalized push tag of the target object is a third preset value, and the activity grouping information and the activation probability are greater than the preset condition threshold, obtain a second resource delivery coefficient for multimedia information delivery; wherein the second resource delivery coefficient is less than the first resource delivery coefficient.

[0153] Step 706: When the personalized push tag of the target object is a fourth preset value, a first resource delivery coefficient for multimedia information delivery is obtained.

[0154] The third preset value and the fourth preset value are only used to distinguish different preset values. For example, Figure 6 As shown in , the third preset value in the present application may be 0, and the fourth preset value may be 1.

[0155] The preset condition threshold refers to the condition threshold used to screen activity grouping information and startup probability. The preset condition threshold in the present application may include a preset activity grouping threshold and a preset startup probability threshold. For example, the preset activity grouping threshold can be set to be less than or equal to the low activity group, that is, when the activity grouping information of the target object is less than or equal to the low activity group, it indicates that the activity of the target object is low; when the activity grouping information of the target object is greater than the low activity group, it indicates that the activity of the target object is high; for another example, the preset startup probability threshold can be set to be less than or equal to 0.5, that is, when the startup probability of the target object is less than or equal to the preset startup probability threshold of 0.5, it indicates that the self-start probability of the target object is low; when the startup probability of the target object is greater than the preset startup probability threshold of 0.5, it indicates that the self-start probability of the target object is high.

[0156] The first resource allocation coefficient and the second resource allocation coefficient are only used to distinguish different types of resource allocation coefficients. For example, the first resource allocation coefficient in this application may be a high resource allocation coefficient, and the second resource allocation coefficient may be a low resource allocation coefficient.

[0157] Specifically, after the terminal determines the target object of the second delivery channel from each object based on the application operation period and the interaction probability, the terminal can determine the resource delivery coefficient of the multimedia information delivery based on the personalized push tag, activity grouping information and activation probability of the target object, for example, Figure 6As shown in, it is a schematic diagram of the overall application framework of multi-dimensional user tags. When the personalized push tag of the target object is 0 (on state), and the activity group information is less than or equal to the low-activity group, and the startup probability is less than or equal to the preset startup probability threshold of 0.5, the first resource delivery coefficient of multimedia information delivery is A1 (high), and A1 is the increased resource delivery coefficient; when the personalized push tag of the target object is 0 (on state), and the activity group information is greater than the low-activity group, and the startup probability is greater than the preset startup probability threshold of 0.5, the second resource delivery coefficient A2 (low) of multimedia information delivery is obtained, and A2 is the reduced resource delivery coefficient; wherein, the second resource delivery coefficient A2 < the first resource delivery coefficient A2; when the personalized push tag of the target object is 1 (off state), the first resource delivery coefficient of multimedia information delivery is A1 (high), and A1 is the increased resource delivery coefficient. As a result, the delivery efficiency of multimedia information can be effectively improved, and the delivery effect of multimedia information can be improved at the same time.

[0158] The present application also provides an application scenario, and the application scenario applies the above-mentioned multimedia information delivery method. Specifically, the application of the multimedia information delivery method in the application scenario is as follows:

[0159] In the process of interaction between the advertiser (multimedia merchant terminal) and the multimedia information platform, the above-mentioned multimedia information delivery method can be used. When the advertiser (multimedia merchant terminal) receives the traffic request information sent by the multimedia information platform, the advertiser (multimedia merchant terminal) responds to the traffic request information, obtains the application operation record, resource interaction record, personalized push tag, application operation time period and activity grouping information of each object in the application in the first delivery channel, and determines the startup probability of each object for the application based on the application operation record; the advertiser (multimedia merchant terminal) determines the interaction probability of each object for resource interaction based on the resource interaction record, and determines the target object of the second delivery channel from each object based on the application operation period and the interaction probability; the advertiser (multimedia merchant terminal) determines the resource delivery coefficient of the multimedia information delivery based on the personalized push tag, activity grouping information and startup probability of the target object; the advertiser (multimedia merchant terminal) delivers the multimedia information of the second delivery channel to the target object based on the resource delivery coefficient. As a result, when advertisers (multimedia merchant terminals) interact with multimedia information platforms, the advertising delivery method provided in this application can effectively improve the ROI of member payments, and as an in-terminal activity method of multimedia merchant terminals, it can effectively improve the delivery efficiency of growth activities and increase the probability of member payments.

[0160] The method provided in the embodiment of the present application can be applied to various multimedia information delivery scenarios. The following takes the scenario where an advertiser (multimedia merchant terminal) interacts with a multimedia information platform as an example to illustrate the multimedia information delivery method provided in the embodiment of the present application.

[0161] oCPA (Optimized Cost per Action) advertising: bid according to optimization goals, charge by click or exposure. When advertisers select a specific optimization goal (for example: activation of mobile applications, ordering on websites) in the advertising delivery process, provide the average price they are willing to pay for this delivery goal, and send back the performance data in a timely and accurate manner, with the help of the conversion estimation model, the conversion value of each click to the advertiser is estimated in real time, and the bid is automatically made, and finally the fee is deducted according to the click; at the same time, the conversion estimation model will be continuously and automatically optimized according to the advertiser's advertising conversion data.

[0162] RTA stands for Real Time API, which can quickly respond to user interest migration by mining massive amounts of data in real time. RTA technology is applicable to multiple industries, and after testing, it can achieve a device duplication rate as low as 2. It can bring advertisers a smarter way of advertising and more hardcore technical support, effectively solve the flexible volume requirements of direct investment customers, and support customers in major vertical industries to perform in-depth effect optimization, thereby achieving efficient customer acquisition. RTA was launched to meet advertisers' needs for efficient use of traffic, and to help advertisers avoid traffic loss and obtain high-quality user conversions more accurately.

[0163] The client, or advertiser, provides the advertising system with subject matter and advertising conversion data.

[0164] pCVR (predicted CVR): predicted conversion rate.

[0165] pCTR: estimated click-through rate.

[0166] With the rise of oCPA+RTA advertising, advertisers can more finely control the amount of advertising, such as targeted user screening, differentiated bidding strategies, and personalized content, to improve the conversion ability of advertising. Figure 4 As shown in , when the advertising platform obtains a traffic access request, it forwards the traffic request information (user device, ad space, etc.) to advertisers that support RTA channels (i.e., RTA request external), and the advertisers decide whether to participate in the platform bidding, that is, whether to display the advertisement under the current traffic, and what the bidding price is.

[0167] For advertisers, they need to pay advertising fees after each successful bid. The higher the ROI (return on investment), the more willing the advertiser is to participate in the platform bidding.

[0168] The technical solutions in the traditional way include:

[0169] (1) Crowd screening based on playback probability or retention probability

[0170] Technical solution: Screen the audience by estimating playback probability, second retention probability, etc.

[0171] Disadvantages: This method is relatively simple and avoids the problem of sparse payment data, reducing the difficulty of model learning. However, since proxy goals such as playback and retention are not necessarily positively correlated with payment indicators, the efficiency of population screening is low and it is impossible to achieve decisions that maximize ROI.

[0172] (2) Crowd screening based on historical statistics

[0173] Technical solution: This method uses users with a history of payment behavior as a whitelist, without the need for model estimation, and directly screens the population from the whitelist.

[0174] Disadvantages: This method has high accuracy in crowd screening, but the crowd screening set is small, and it is impossible to expand the scale of advertising to improve the advertising volume.

[0175] The traditional technical solution has the following disadvantages:

[0176] (1) Modeling through proxy goals (playback / retention, etc.) cannot directly optimize ROI, and the delivery effect is not ideal.

[0177] (2) Based on historical statistical methods, the population set is too small to achieve scaled advertising.

[0178] (3) Data modeling that only considers a single scenario cannot achieve overall business linkage.

[0179] Therefore, in order to solve the above problems, this application provides an RTA crowd selection solution based on self-starting delivery linkage. For the buying scenario of member-paid applications, this application proposes an advertising delivery method that is linked to its own startup behavior. It uses a deep learning model to estimate the user's payment probability, and combines the user's own activity and activity in free channels to estimate the user's payment behavior, and bids based on this to improve the ROI of advertising delivery.

[0180] The key technologies in the technical solution of this application are as follows:

[0181] 1. Multi-objective learning technology. Due to the various forms of paid content, such as membership renewal, additional on-demand payment, and membership payment by ordinary users, a multi-objective model of active payment + automatic payment is built to estimate the probability value of each payment type.

[0182] 2. Multi-dimensional object feature characterization. Based on the user's historical behavior, the user's self-start probability, push click activity, RTA payment probability and other features are constructed through statistical methods, deep learning models and other methods.

[0183] 3. Overall model application framework. Formulate model application rules based on multi-dimensional object features, implement crowd value screening and advertising bidding at the same time, and realize one-stop purchase volume decision-making.

[0184] On the technical side,

[0185] 1.1 Technical Goals

[0186] The technical solution of the present application uses a deep learning model to model the payment probability and user self-start probability to identify the user's delivery value and self-start probability, and combines multi-dimensional features such as user click activity to comprehensively screen the delivery population to retain high-value users, and set the bidding coefficients of people in different circles based on the payment probability and start probability to achieve the goal of delivery ROI>=1.

[0187] 1.2 Advertising delivery based on self-starting delivery linkage

[0188] (1) Overall architecture

[0189] like Figure 8 As shown, it is a schematic diagram of the algorithm architecture of the advertising delivery method linked to the self-starting behavior. Figure 8 In the algorithm architecture diagram of the technical solution provided by the present application, the overall architecture consists of three parts: multi-objective learning, multi-dimensional object feature characterization, and model application.

[0190] Among them, 1) the multi-objective learning module is responsible for the training of the payment probability model. Considering that the behavior patterns of members and non-members may be different, in the payment scenario, the object features of prediction of active payment and automatic payment of users will also be different. Therefore, the model objectives are set as active payment probability estimation, automatic payment probability estimation, and total payment probability estimation.

[0191] 2) Multi-dimensional object feature characterization, which is responsible for characterizing users from the dimensions of user payment probability, self-initiation probability, click activity, etc. Specifically, the clustering label method can be used to characterize the user's click activity in the push channel (free channel), and the machine learning / deep learning method can be used to characterize the probability prediction of user self-initiation and the payment probability prediction in the advertising channel.

[0192] 3) Model application module, comprehensively use multi-dimensional object features, formulate business application rules, realize RTA channel population screening, and set different bid coefficients for released populations. For users with high probability of self-startup or high probability of pulling up in free channels (such as push channels), reduce the release of RTA channels or lower the bid coefficient. Conversely, increase the release of RTA channels or increase the bid coefficient.

[0193] (2) Multi-objective learning model

[0194] Since member payment is a renewal behavior and non-member payment is a first-time payment behavior, the two have significant differences in behavior. In model training, they can be learned as two tasks, that is, the automatic renewal task (binary classification, whether to renew) and the active payment task (binary classification, whether to pay for the first time) are learned at the same time. The automatic renewal samples only affect the network of the automatic renewal probability part, and the active payment samples only affect the network of the active payment probability part. The two probabilities are added together as the output of the overall payment probability.

[0195] In terms of network structure, the technical solution provided in this application can adopt a PLE network structure, such as Figure 5 As shown in . In order to further solve the problem of scarce positive samples, the model loss function adopts focal loss as shown in the above formula (1), which is modified based on the standard cross entropy loss. This loss function can reduce the weight of easy-to-classify samples, so that the model can focus more on difficult-to-classify samples during training.

[0196] (3) Characterization of multidimensional objects

[0197] User click activity grouping

[0198] According to the number of push messages received and clicked messages in the push channel (free push channel) in the past month, users can be divided into the following categories: Figure 3 As shown in the table in Figure 3, the activity of the crowd increases from top to bottom.

[0199] Personalized push switch label

[0200] Whether the user turns on the personalized push switch in the application is a fact label. If the switch is turned on, the user can be reached through the push channel (free channel), otherwise he cannot be reached.

[0201] User start probability prediction

[0202] Based on the user's historical activity records, the model is trained through machine learning / deep learning methods to predict the probability of users actively opening the application the next day. The training target is binary classification. After the model training is completed, a score can be given to a given user group. The value range is [0, 1]. The higher the prediction score, the higher the probability that the user will actively launch the application.

[0203] User payment probability prediction

[0204] Based on the user's historical payment records, the model is trained using deep learning methods to predict the probability that the user will pay after being pulled in by the RTA channel the next day. The value range is [0, 1]. The higher the prediction score, the higher the probability that the user will pay.

[0205] (4) Model application method

[0206] Based on the above multi-dimensional object feature characterization and combined with business scenarios, an application system was developed to control the user release of the RTA channel and set the bid coefficients of different users. The specific application principles are as follows:

[0207] For users who can be reached by free channels (such as push channels) and have a high probability of being pulled in, appropriately reduce the amount of users released by paid channels (RTA channels) or lower the bid coefficient;

[0208] For users with a high probability of activation, reduce the amount of paid channels (RTA channels) released, or lower the bid coefficient;

[0209] For users who cannot be reached by free channels (such as push channels) and have a low activation probability, increase the release volume of paid channels (RTA channels) or increase the bid coefficient;

[0210] For users with high payment probability prediction values, increase the amount of payment channels (RTA channels) released and increase the bid coefficient; for users with low payment probability, reduce the amount of payment channels (RTA channels) released and lower the bid coefficient. The specific release rules are as follows: Figure 6 as shown in .

[0211] The beneficial effects of the technical solution of this application include:

[0212] (1) As a method of advertising, it can improve the ROI of paid membership applications.

[0213] (2) As a method of on-end activities, it can not only improve the efficiency of growth activities, but also effectively increase the probability of members paying.

[0214] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0215] Based on the same inventive concept, the embodiment of the present application also provides a multimedia information delivery device for implementing the multimedia information delivery method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of one or more multimedia information delivery devices provided below can refer to the limitations of the multimedia information delivery method above, and will not be repeated here.

[0216] In one embodiment, Fig. 9 As shown, a multimedia information delivery device is provided, including: an acquisition module 902, a determination module 904 and a delivery module 906, wherein:

[0217] The acquisition module 902 is used to acquire application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in the application program in the first delivery channel.

[0218] Determination module 904 is used to determine the probability of each object starting the application based on the application operation record; determine the interaction probability of each object performing resource interaction based on the resource interaction record; determine the target object of the second delivery channel from each object based on the application operation period and the interaction probability; determine the resource delivery coefficient of the multimedia information delivery based on the personalized push tag, activity grouping information and startup probability of the target object.

[0219] The delivery module 906 is used to deliver the multimedia information of the second delivery channel to the target object based on the resource delivery coefficient.

[0220] In one embodiment, the device also includes: a sending module and a receiving module, the sending module is used to send the object information of the target object and the resource delivery coefficient to the multimedia information platform, so that the multimedia information platform compares the resource delivery coefficient with other resource delivery coefficients and determines the result data based on the comparison result; the receiving module is used to receive the result data returned by the multimedia information platform; the delivery module is also used to deliver multimedia information of the second delivery channel to the target object based on the resource delivery coefficient when the result data meets the preset conditions.

[0221] In one embodiment, the device also includes: a receiving module, used to receive traffic request information sent by a multimedia information platform; the traffic request information is sent by the multimedia information platform to a client that supports the second delivery channel; the acquisition module is also used to respond to the traffic request information, obtain application operation records and resource interaction records of each object in the application, and obtain personalized push tags for each object and activity grouping information in the first delivery channel; the determination module is also used to determine the application operation time period based on the application operation records.

[0222] In one embodiment, the acquisition module is also used to obtain personalized push tags for each object in the application; obtain the number of push information and the number of click information for each of the objects in the first delivery channel; the determination module is also used to determine the activity grouping information of each of the objects in the first delivery channel based on the number of push information, the number of click information and time information.

[0223] In one embodiment, the determination module is also used to determine that the object is in the first active group in the first delivery channel when the time information is a first preset time period and the number of click information is greater than a first preset value and not greater than a second preset value; to determine that the object is in the second active group in the first delivery channel when the time information is a second preset time period and the number of click information is greater than the first preset value and not greater than the second preset value; and to determine that the object is in the third active group in the first delivery channel when the time information is the second preset time period and the number of click information is greater than the second preset value; wherein the activity levels of the first active group, the second active group and the third active group increase in sequence.

[0224] In one embodiment, the acquisition module is also used to obtain the activity grouping information of each of the objects in the first delivery channel; the device also includes: a detection module, used to detect the on / off status of the personalized push function of each of the objects in the application; the determination module is also used to determine the personalized push label of each of the objects based on the on / off status.

[0225] In one embodiment, the acquisition module is further used to acquire the target time from the application operation record; the determination module is further used to determine the difference between the target time and the current time; and the difference is used as the application operation period of each object.

[0226] In one embodiment, the device also includes: a processing module, which is used to classify the resource interaction records of each object through each network model in the classification model to obtain a first resource interaction probability and a second resource interaction probability; the determination module is also used to determine the interaction probability of each object performing resource interaction based on the first resource interaction probability and the second resource interaction probability.

[0227] In one embodiment, the acquisition module is also used to acquire training samples including the first resource interaction and the second resource interaction; the device also includes: a training module, used to train the first network model and the second network model in the initial classification model based on the training samples until the sum of the first loss value of the first network model and the second loss value of the second network model meets the loss value condition, thereby obtaining the classification model.

[0228] In one embodiment, the determination module is also used to, when the application operation period is less than or equal to a preset time threshold and the interaction probability is greater than a preset probability value, select the object among the objects whose application operation period is less than or equal to the preset time threshold and whose interaction probability is greater than the preset probability value as the target object of the second delivery channel.

[0229] In one embodiment, the device also includes: an acquisition module, which is used to obtain a first resource delivery coefficient for multimedia information delivery when the personalized push tag of the target object is a third preset value, and the activity grouping information and the startup probability are less than or equal to a preset condition threshold; when the personalized push tag of the target object is the third preset value, and the activity grouping information and the startup probability are greater than the preset condition threshold, obtain a second resource delivery coefficient for multimedia information delivery; wherein the second resource delivery coefficient is less than the first resource delivery coefficient; when the personalized push tag of the target object is a fourth preset value, obtain the first resource delivery coefficient for multimedia information delivery.

[0230] Each module in the multimedia information delivery device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0231] In one embodiment, a computer device is provided. The computer device may be a terminal or a server. In this embodiment, the computer device is described as a terminal. The internal structure diagram thereof may be as follows: Fig.10 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for delivering multimedia information is implemented. The display unit of the computer device is used to form a visually visible image, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse, etc.

[0232] Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0233] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0234] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0235] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0236] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0237] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0238] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0239] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for delivering multimedia information, characterized in that: The method comprises: Obtain application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in the application program in the first delivery channel; Determining the probability of each of the objects starting the application program based on the application operation record; Determining the interaction probability of each of the objects performing resource interaction based on the resource interaction record; Determining a target object of a second delivery channel from each of the objects based on the application operation period and the interaction probability; Determining a resource delivery coefficient for multimedia information delivery based on the personalized push tag of the target object, the activity grouping information, and the activation probability; Based on the resource delivery coefficient, multimedia information of the second delivery channel is delivered to the target object.

2. The method according to claim 1, characterized in that: The method further comprises: Sending the object information of the target object and the resource delivery coefficient to the multimedia information platform, so that the multimedia information platform compares the resource delivery coefficient with other resource delivery coefficients, and determines result data based on the comparison result; Receiving the result data returned by the multimedia information platform; The delivering the multimedia information of the second delivery channel to the target object based on the resource delivery coefficient includes: When the result data meets the preset condition, multimedia information delivery through the second delivery channel is performed to the target object based on the resource delivery coefficient.

3. The method according to claim 1, characterized in that: The method further comprises: receiving flow request information sent by the multimedia information platform; the flow request information is sent by the multimedia information platform to the client supporting the second delivery channel; The obtaining of application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in the application program in the first delivery channel includes: In response to the traffic request information, obtaining application operation records and resource interaction records of each object in the application, and obtaining personalized push tags and activity grouping information of each object in the first delivery channel; An application operation period is determined according to the application operation record.

4. The method according to claim 3, characterized in that The obtaining of the personalized push tag of each object and the activity grouping information in the first delivery channel includes: Get personalized push tags for each object in the application; Obtain the number of pushed information and clicked information of each object in the first delivery channel; Based on the number of push information, the number of click information and the time information, activity grouping information of each of the objects in the first delivery channel is determined.

5. The method according to claim 4, characterized in that The determining, based on the number of push information, the number of click information and the time information, the activity grouping information of each object in the first delivery channel includes: When the time information is a first preset time period and the number of click information is greater than a first preset value and not greater than a second preset value, it is determined that the object is in a first active group in the first delivery channel; When the time information is a second preset time period and the number of click information is greater than the first preset value and not greater than the second preset value, it is determined that the object is in the second active group in the first delivery channel; When the time information is the second preset time period and the number of click information is greater than a second preset value, it is determined that the object is in the third active group in the first delivery channel; The activity levels of the first active group, the second active group, and the third active group increase in sequence.

6. The method according to claim 3, characterized in that The obtaining of the personalized push tag of each object and the activity grouping information in the first delivery channel includes: Obtaining activity grouping information of each of the objects in the first delivery channel; Detecting the on / off status of the personalized push function of each object in the application; A personalized push tag for each of the objects is determined according to the open and closed states.

7. The method according to claim 3, characterized in that The determining the application operation period according to the application operation record includes: Obtaining a target time from the application operation record; determining a difference between the target time and the current time; The difference is used as the application operation period of each of the objects.

8. The method according to claim 1, characterized in that The determining, based on the resource interaction record, the interaction probability of each object performing resource interaction includes: Classify the resource interaction records of each object through each network model in the classification model to obtain a first resource interaction probability and a second resource interaction probability; The interaction probability of each of the objects performing resource interaction is determined according to the first resource interaction probability and the second resource interaction probability.

9. The method according to claim 8, characterized in that Before classifying the resource interaction records of each object by using each network model in the classification model to obtain the first resource interaction probability and the second resource interaction probability, the method further includes: Acquire a training sample including a first resource interaction and a second resource interaction; The first network model and the second network model in the initial classification model are trained based on the training samples until the sum of the first loss value of the first network model and the second loss value of the second network model meets the loss value condition, thereby obtaining the classification model.

10. The method according to claim 1, characterized in that The determining the target object of the second delivery channel from each of the objects based on the application operation period and the interaction probability includes: When the application operation period is less than or equal to the preset time threshold and the interaction probability is greater than the preset probability value, the object among the objects whose application operation period is less than or equal to the preset time threshold and whose interaction probability is greater than the preset probability value is taken as the target object of the second delivery channel.

11. The method according to claim 1, characterized in that: The determining of a resource delivery coefficient for multimedia information delivery based on the personalized push tag of the target object, the activity grouping information, and the activation probability includes: When the personalized push tag of the target object is a third preset value, and the activity grouping information and the activation probability are less than or equal to a preset condition threshold, obtaining a first resource delivery coefficient for multimedia information delivery; When the personalized push tag of the target object is the third preset value, and the activity grouping information and the activation probability are greater than the preset condition threshold, a second resource delivery coefficient for multimedia information delivery is obtained; wherein the second resource delivery coefficient is less than the first resource delivery coefficient; When the personalized push tag of the target object is a fourth preset value, the first resource delivery coefficient for multimedia information delivery is obtained.

12. A multimedia information delivery device, characterized in that: The device comprises: An acquisition module is used to acquire application operation records, resource interaction records, personalized push tags, application operation time periods, and activity grouping information of each object in the application program in the first delivery channel; A determination module, configured to determine the probability of each of the objects starting the application based on the application operation record; determine the probability of each of the objects interacting with resources based on the resource interaction record; determine the target object of the second delivery channel from each of the objects based on the application operation period and the interaction probability; determine the resource delivery coefficient for multimedia information delivery based on the personalized push tag, the activity grouping information and the launch probability of the target object; A delivery module is used to deliver the multimedia information of the second delivery channel to the target object based on the resource delivery coefficient.

13. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 11 are implemented.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

15. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images