A data processing method, device, equipment and readable storage medium

Abstract

This application discloses a data processing method, apparatus, device, and readable storage medium. The method includes: acquiring user environment information corresponding to a target user and browsing behavior-related information of the target user regarding first public broadcast media data within a first time period; determining user browsing status characteristics of the target user regarding the first public broadcast media data based on the user environment information and browsing behavior-related information; predicting a first predicted quantity corresponding to the target user based on the user browsing status characteristics; determining a second predicted quantity based on the first predicted quantity; and pushing second public broadcast media data to the terminal corresponding to the target user within a second time period based on the first predicted quantity and the second predicted quantity. Using this application can improve the accuracy of media data type recommendation.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a data processing method, apparatus, device, and readable storage medium. Background Technology

[0002] With the development of multimedia technology, different types of multimedia data (such as video, text, and music) have become the main carriers for people to obtain information and enjoy entertainment in their daily lives. Currently, multimedia data recommendation often requires recommending multiple types of multimedia data to users simultaneously. However, different users have different preferences for different types of multimedia data. Therefore, how to accurately recommend multimedia data that matches users' preferences has become a very important issue.

[0003] Currently, the main approach involves offline calculation of users' interests (preference levels) for different types of multimedia data, and recommending more multimedia data of that type (e.g., video) to users who prefer to browse a certain type. However, this method of offline calculation of user interests is slow to update. Over time, the results determined by offline calculation may diverge from the user's interests, ultimately leading to inaccurate multimedia data recommendations that do not match the user's interests. Summary of the Invention

[0004] This application provides a data processing method, apparatus, device, and readable storage medium that can improve the accuracy of media data type recommendation.

[0005] One embodiment of this application provides a data processing method, including:

[0006] Within the first time period, acquire user environment information corresponding to the target user, as well as information related to the target user's browsing behavior on the first public broadcast media data;

[0007] Based on user environment information and browsing behavior-related information, determine the user browsing status characteristics of the target user for the first public broadcast media data;

[0008] The first predicted quantity is predicted based on the user's browsing status characteristics; the first predicted quantity refers to the number of media data belonging to the first media type pushed to the terminal corresponding to the target user during the second time period; the first time period is earlier than the second time period.

[0009] The second predicted quantity is determined based on the first predicted quantity; the second predicted quantity refers to the amount of media data belonging to the second media type pushed to the terminal corresponding to the target user within the second time period.

[0010] Based on the first predicted quantity and the second predicted quantity, the second public broadcast media data is pushed to the terminal corresponding to the target user within the second time period; the second public broadcast media data includes the first target media data belonging to the first media type and the second target media data belonging to the second media type; the quantity of the first target media data is the first predicted quantity; the quantity of the second target media data is the second predicted quantity.

[0011] One embodiment of this application provides a data processing apparatus, including:

[0012] The information acquisition module is used to acquire user environment information corresponding to the target user and browsing behavior related information of the target user for the first public broadcast media data within the first time period.

[0013] The feature determination module is used to determine the user browsing status features of the target user in relation to the first public broadcast media data based on user environment information and browsing behavior-related information.

[0014] The quantity determination module is used to predict the first predicted quantity corresponding to the target user based on the user's browsing status characteristics; the first predicted quantity refers to the quantity of media data belonging to the first media type pushed to the terminal corresponding to the target user within the second time period; the first time period is earlier than the second time period;

[0015] The quantity determination module is also used to determine a second predicted quantity based on the first predicted quantity; the second predicted quantity refers to the quantity of media data belonging to the second media type pushed to the terminal corresponding to the target user within the second time period.

[0016] The data push module is used to push the second public broadcast media data to the terminal corresponding to the target user within a second time period based on the first predicted quantity and the second predicted quantity; the second public broadcast media data includes the first target media data belonging to the first media type and the second target media data belonging to the second media type; the quantity of the first target media data is the first predicted quantity; the quantity of the second target media data is the second predicted quantity.

[0017] In one embodiment, user environment information includes user attribute information and environment attribute information;

[0018] The feature determination module includes:

[0019] The hash value generation unit is used to generate the first hash value corresponding to the user attribute information based on the user attribute information and the hash function.

[0020] The hash value generation unit is also used to generate a second hash value corresponding to the environmental attribute information based on the environmental attribute information and the hash function;

[0021] The hash value generation unit is also used to generate a third hash value corresponding to the browsing behavior association information based on the browsing behavior association information and the hash function;

[0022] The feature determination unit is used to determine the user browsing status features of the target user in relation to the first public broadcast media data based on the first hash value, the second hash value, and the third hash value.

[0023] In one embodiment, the feature determination unit includes:

[0024] The table retrieval sub-unit is used to retrieve the vector mapping table; the vector mapping table includes the mapping relationship between configuration hash values ​​and configuration vectors; the configuration hash values ​​include a first hash value, a second hash value, and a third hash value;

[0025] The vector acquisition subunit is used to retrieve a configuration vector that has a mapping relationship with the first hash value from the vector mapping table, and use it as the first user feature vector;

[0026] The vector acquisition subunit is also used to retrieve a configuration vector that has a mapping relationship with the second hash value from the vector mapping table, as the second user feature vector;

[0027] The vector acquisition subunit is also used to retrieve a configuration vector that has a mapping relationship with the third hash value from the vector mapping table, as the third user feature vector;

[0028] The vector fusion subunit is used to fuse the first user feature vector, the second user feature vector, and the third user feature vector to obtain the user browsing status features of the target user in relation to the first public broadcast media data.

[0029] In one embodiment, the quantity determination module includes:

[0030] The probability prediction unit is used to input user browsing status features into the target quantity prediction model, and output the prediction probabilities corresponding to N configured quantities through the target quantity prediction model to obtain N prediction probabilities; each configured quantity represents the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user in the second time period; N is an integer;

[0031] The first quantity determination unit is used to obtain the maximum prediction probability among N prediction probabilities and determine the configuration quantity corresponding to the maximum prediction probability as the first prediction quantity.

[0032] In one embodiment, the quantity determination module includes:

[0033] The quantity acquisition unit is used to acquire the total configuration quantity; the total configuration quantity refers to the total amount of media data required to push to the terminal corresponding to the target user.

[0034] The second quantity determination unit is used to subtract the total configuration quantity from the first predicted quantity to obtain the second predicted quantity.

[0035] In one embodiment, the data push module includes:

[0036] The collection acquisition unit is used to acquire a collection of media data to be pushed from the database. The collection of media data to be pushed includes media data of the first media type that is expected to be pushed to the terminal corresponding to the target user during the second time period, and media data of the second media type that is expected to be pushed to the terminal corresponding to the target user.

[0037] The push data acquisition unit is used to acquire the first target media data from the set of media data to be pushed based on the first predicted quantity;

[0038] The push data acquisition unit is also used to acquire second target media data from the set of media data to be pushed based on the second predicted quantity;

[0039] The data push unit is used to identify the first target media data and the second target media data as the second public broadcast media data, and push the second public broadcast media data to the terminal corresponding to the target user within the second time period.

[0040] This application provides another data processing method, including:

[0041] The browsing status features of a first sample user and a second sample user associated with the target user are obtained. The browsing status features of the first sample user are determined based on the target user's first sample user environment information obtained within the first sample time period, and the target user's first browsing behavior related information regarding the first sample public broadcast media data. The browsing status features of the second sample user are determined based on the target user's second sample user environment information obtained within the second sample time period, and the target user's second browsing behavior related information regarding the second sample public broadcast media data. The first sample time period is earlier than the second sample time period.

[0042] Based on the browsing status characteristics of the first sample user and the browsing status characteristics of the second sample user, training sample data associated with the target user is determined.

[0043] The quantity prediction model is trained based on the training sample data to obtain the target quantity prediction model. The target quantity prediction model is used to predict the number of types corresponding to the browsing status characteristics of the target user. The number of types refers to the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user.

[0044] One embodiment of this application provides another data processing apparatus, including:

[0045] The feature acquisition module is used to acquire the browsing status features of a first sample user and a second sample user associated with the target user. The browsing status features of the first sample user are determined based on the target user's first sample user environment information acquired within the first sample time period, and the target user's first browsing behavior related information regarding the first sample public broadcast media data. The browsing status features of the second sample user are determined based on the target user's second sample user environment information acquired within the second sample time period, and the target user's second browsing behavior related information regarding the second sample public broadcast media data. The first sample time period is earlier than the second sample time period.

[0046] The training sample determination module is used to determine the training sample data associated with the target user based on the browsing status characteristics of the first sample user and the browsing status characteristics of the second sample user.

[0047] The model training module is used to train the quantity prediction model based on the training sample data to obtain the target quantity prediction model. The target quantity prediction model is used to predict the number of types corresponding to the browsing status characteristics of the target user. The number of types refers to the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user.

[0048] In one embodiment, the feature acquisition module includes:

[0049] The sample information acquisition unit is used to acquire, within the first sample time period, the first sample user environment information corresponding to the target user, and the first sample browsing behavior association information of the target user for the first sample public broadcast media data.

[0050] The sample state determination unit is used to determine the first sample user browsing state characteristics of the target user based on the first sample user environment information and the first sample browsing behavior association information.

[0051] The sample quantity determination unit is used to select the target configuration quantity from N configuration quantities as the first sample type quantity corresponding to the first sample user browsing status characteristics; the first sample type quantity refers to the quantity of media data belonging to the first media type pushed to the terminal corresponding to the target user during the second sample time period;

[0052] The sample quantity determination unit is also used to obtain the total configuration quantity; the total configuration quantity refers to the total amount of media data required to push to the terminal corresponding to the target user.

[0053] The sample feature determination unit is used to determine the browsing status features of the second sample user based on the total number of configurations and the number of first sample types.

[0054] In one embodiment, the sample size determination unit includes:

[0055] The sample quantity determination subunit is used to subtract the total number of configurations from the number of the first sample type to obtain the number of the second sample type of media data belonging to the second media type pushed to the terminal corresponding to the target user during the second sample time period.

[0056] The sample data push subunit is used to push the second sample public broadcast media data to the terminal corresponding to the target user within the second sample time period according to the number of the first sample type and the number of the second sample type; the second sample public broadcast media data includes the first sample media data belonging to the first media type and the second sample media data belonging to the second media type; the number of the first sample media data is the number of the first sample type, and the number of the second sample media data is the number of the second sample type;

[0057] The sample data acquisition subunit is used to acquire the second sample user environment information corresponding to the target user and the second sample browsing behavior association information of the target user for the second sample public broadcast media data within the second sample time period.

[0058] The sample feature determination subunit is used to determine the second sample user state features of the target user based on the second sample user environment information and the second sample browsing behavior association information.

[0059] In one embodiment, the second sample browsing behavior association information includes the sample click frequency of the target user on the second sample public broadcast media data;

[0060] The training sample determination module includes:

[0061] The parameter determination unit is used to determine the sample click frequency as a feedback parameter associated with the target user and the number of the first sample type;

[0062] The training sample determination unit is used to determine the data group consisting of the browsing status features of the first sample user, the number of first sample types, feedback parameters, and the browsing status features of the second sample user as training sample data associated with the target user.

[0063] In one embodiment, the model training module includes:

[0064] The feature input unit is used to input the first sample user browsing status feature from the training sample data into the quantity prediction model. The quantity prediction model and the first sample user browsing status feature output the first sample prediction probability corresponding to the N configuration video quantities, respectively, to obtain N first sample prediction probabilities.

[0065] The probability selection unit is used to determine the first sample prediction probability corresponding to the number of first sample types in the training sample data from N first sample prediction probabilities, and use it as the first probability to be calculated.

[0066] The feature input unit is also used to input the second sample user browsing status features in the training sample data into the quantity prediction model, and output the second sample prediction probabilities corresponding to N configuration quantities through the quantity prediction model and the second sample user browsing status features, so as to obtain N second sample prediction probabilities.

[0067] The optimal quantity determination unit is used to obtain the maximum second sample prediction probability among N second sample prediction probabilities, and determine the configuration quantity corresponding to the maximum second sample prediction probability as the optimal type quantity.

[0068] The feature input unit is also used to input the browsing status features of the second sample user into the quantity evaluation model, and output the prediction probabilities of the third sample corresponding to N configuration quantities through the quantity evaluation model and the browsing status features of the second sample user, so as to obtain N prediction probabilities of the third sample.

[0069] The model training unit is used to train the quantity prediction model based on the first probability to be calculated, the optimal number of types, the feedback parameters in the training sample data, and the prediction probabilities of N third samples, so as to obtain the target quantity prediction model.

[0070] In one embodiment, the model training unit includes:

[0071] The probability acquisition subunit is used to obtain the third sample prediction probability corresponding to the optimal number of types from N third sample prediction probabilities, and use it as the second probability to be calculated.

[0072] The probability operation subunit is used to obtain the discount coefficient of the quantity prediction model, and determine the target operation probability associated with the training sample data based on the discount coefficient, feedback parameters, and the second probability to be operated.

[0073] The loss value generation subunit is used to subtract the target operation probability from the first operation probability to obtain the operation result, and determine the loss function value of the quantity prediction model based on the operation result;

[0074] The model training subunit is used to train the quantity prediction model based on the loss function value to obtain the target quantity prediction model.

[0075] One embodiment of this application provides a computer device, including: a processor and a memory;

[0076] The memory stores a computer program, which, when executed by a processor, causes the processor to perform the methods described in the embodiments of this application.

[0077] One aspect of this application provides a computer-readable storage medium storing a computer program, which includes program instructions. When executed by a processor, the program instructions perform the methods described in this application.

[0078] One aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in one aspect of the embodiments of this application.

[0079] In this embodiment, user environment information corresponding to a target user within a first time period and browsing behavior association information of the target user regarding the first public broadcast media data can be obtained. Based on the user environment information and browsing behavior association information, the user browsing status characteristics of the target user (i.e., the target user's current preference status characteristics for different types of media data) can be determined. Based on the user browsing status characteristics, a first predicted quantity corresponding to the target user can be predicted. Based on the first predicted quantity, a second predicted quantity can be determined. Based on the first and second predicted quantities, the second public broadcast media data can be pushed to the target user within the second time period. The second public broadcast media data includes media data belonging to a first media type (i.e., the first target media data) and media data belonging to a second media type (i.e., the second target media data). The quantity of the first target media data is the first predicted quantity, and the quantity of the second target media data is the second predicted quantity. In other words, this application can predict, based on the user browsing status characteristics of the target user in the first time period, the number of media data belonging to the first media type and the number of media data belonging to the second media type when pushing second public broadcast media data to the target user in the second time period. Since the number of media data belonging to the first media type and the number of media data belonging to the second media type are predicted and determined based on the user browsing status characteristics (current preference status characteristics) of the target user, the number of media data belonging to the first media type and the number of media data belonging to the second media type are consistent with the target user's current preference status for the first and second media types, resulting in high accuracy. In summary, this application can improve the accuracy of media data type recommendation. Attached Figure Description

[0080] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0081] Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application;

[0082] Figure 2 This is a schematic diagram illustrating a scenario for pushing media data, provided in an embodiment of this application.

[0083] Figure 3 This is a flowchart illustrating a data processing method provided in an embodiment of this application;

[0084] Figure 4 This is an architecture diagram for training and applying a model, provided in an embodiment of this application;

[0085] Figure 5 This is a schematic diagram illustrating the push of media data provided in an embodiment of this application;

[0086] Figure 6 This is a flowchart illustrating another data processing method provided in an embodiment of this application;

[0087] Figure 7 This is a schematic diagram of a model training architecture provided in an embodiment of this application;

[0088] Figure 8 This is a schematic diagram of the structure of a data processing device provided in an embodiment of this application;

[0089] Figure 9 This is a schematic diagram of another data processing device provided in an embodiment of this application;

[0090] Figure 10 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0091] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0092] This application relates to the field of artificial intelligence. For ease of understanding, the following will explain the concepts of artificial intelligence and related technologies.

[0093] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0094] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0095] The solutions provided in this application belong to the field of machine learning (ML) within the field of artificial intelligence.

[0096] Machine Learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0097] Please see Figure 1 , Figure 1 This is a schematic diagram of a network architecture provided in an embodiment of this application. For example... Figure 1 As shown, this network architecture may include a service server 1000 and a user terminal cluster. The user terminal cluster may include one or more user terminals; the number of user terminals is not limited here. Figure 1 As shown, multiple user terminals may include user terminal 100a, user terminal 100b, user terminal 100c, ..., user terminal 100n; as Figure 1As shown, user terminals 100a, 100b, 100c, ..., 100n can each connect to the business server 1000 via the network, so that each user terminal can interact with the business server 1000 through the network connection.

[0098] It is understandable that, such as Figure 1 Each user terminal shown can have the target application installed. When the target application runs on each user terminal, it can interact with... Figure 1 The business servers 1000 shown interact with each other, enabling each business server 1000 to receive business data from each user terminal. The target application can include applications with data processing capabilities such as displaying text, images, audio, and video. For example, the application can be a multimedia application (e.g., video push application, text push application, music push application, etc.), an entertainment application (e.g., game application), or any application with multimedia data (e.g., video, text, music) loading and playback capabilities, which can be used by users to play and view multimedia data. This application can be a standalone application or an embedded application integrated into a client (e.g., social client, educational client, and multimedia client, etc.), and is not limited here.

[0099] like Figure 1 As shown, the business server 1000 in this embodiment can be the server corresponding to the target application. The business server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0100] For ease of understanding, the embodiments of this application may be described in detail below. Figure 1 Among the multiple user terminals shown, one user terminal is selected as the target user terminal. This user terminal may include, but is not limited to, smart terminals with multimedia data processing functions (e.g., video playback, music playback, text playback) such as smartphones, tablets, laptops, desktop computers, smart TVs, smart speakers, desktop computers, and smartwatches. For example, embodiments of this application may use… Figure 1 The user terminal 100a shown serves as the target user terminal, which may integrate a target application with the data processing function. In this case, the target user terminal can interact with the business server 1000 through the business data platform corresponding to the target application.

[0101] For example, when a target user uses a target application (e.g., an information push application) on their terminal, the target user can view multimedia data pushed to them within that application. This multimedia data can be referred to as public broadcast media data. It should be understood that public broadcast media data may include media data belonging to a first media type (media data being multimedia data) and media data belonging to a second media type. The target user can click on any media data under any media type to play or watch the public broadcast media data. For example, the public broadcast media data may include media data 1 and 2 belonging to the first media type, and media data 3 belonging to the second media type. The target user can click on media data 1, media data 2, or media data 3 belonging to the second media type to play or watch the public broadcast media data. It should be understood that, taking the multimedia data pushed to the target user as the first public broadcast media data as an example, within a certain period, the business server 1000 can obtain user environment information and browsing behavior association information related to the target user and the first public broadcast media data through the target application. The aforementioned user environment information may include user attribute information (e.g., the target user's age, gender, and geographic location) and environmental attribute information (e.g., the media data push time for the first public broadcast media data, the version information of the target application, the operating system information of the target user's terminal, network status parameters, and playback attribute information such as whether the multimedia data plays automatically). The browsing behavior association information may be the target user's click behavior association information regarding the first public broadcast media data (e.g., the total number of first public broadcast media data pushed to the target user, the number of multimedia data belonging to the first media type within the first public broadcast media data, the total number of clicks by the target user on the first public broadcast media data, and the number of clicks by the target user on multimedia data belonging to the first media type within the first public broadcast media data, etc.). It should be understood that this user environment information and browsing behavior association information can serve as business data associated with the target user obtained by the business server 1000 through the target application.

[0102] Understandably, after acquiring business data (including user environment information and browsing behavior association information corresponding to the target user), the business server 1000 can determine the user browsing status characteristics of the target user for the first public broadcast media data based on the user environment information and browsing behavior association information corresponding to the target user; through the user browsing status characteristics, a first predicted quantity (i.e., the quantity of media data belonging to the first media type to be pushed to the target user terminal corresponding to the target user in the next period of time) can be predicted for the target user; a total configured quantity (i.e., the total quantity of media data required to be pushed to the target user terminal corresponding to the target user) can be obtained; based on the total configured quantity and the first predicted quantity, a second predicted quantity (i.e., the quantity of media data belonging to the second media type to be pushed to the target user terminal corresponding to the target user in the next period of time) can be determined for the target user; based on the first predicted quantity, the business server 1000 can obtain the first media... The system firstly identifies target media data of a certain type and secondly identifies target media data of a certain type (the quantity of the first target media data is the first predicted quantity, and the quantity of the second target media data is the second predicted quantity). Subsequently, the service server 1000 can use the media data composed of the first target media data and the second target media data as the next batch of public broadcast media data to be pushed to the target user terminal (which can be referred to as the second public broadcast media data). The service server 1000 can return the second public broadcast media data, which includes the first target media data and the second target media data, to the target user terminal. When the next media data push time arrives (the next media data push time is within the next period of time), the service server 1000 can present the second public broadcast media data, which includes the first target media data and the second target media data, in the target application. The target user can view the first target media data and the second target media data in the display interface of the target user terminal.

[0103] It is understood that both the first and second media types mentioned above can include video, music, text, etc. When the first media type is a certain media type, the second media type can be any other media type besides that type. For example, when the first media type is video, the second media type can be any other media type besides video, such as music, text, etc.

[0104] Optionally, it is understood that the aforementioned user terminal can obtain business data (such as user environment information and browsing behavior association information corresponding to the target user). After obtaining the business data, the user terminal can determine the user browsing status characteristics of the target user for public broadcast media data, predict the first predicted quantity and the second predicted quantity corresponding to the target user through the user browsing status characteristics, obtain the first target media data in the quantity of the first predicted quantity and the second target media data in the quantity of the second predicted quantity, and push the first target media data and the second target media data to the target user.

[0105] It is understood that the methods provided in this application embodiment can be executed by computer equipment, including but not limited to user terminals or business servers. The user terminals and business servers can be directly or indirectly connected via wired or wireless communication, and this application does not impose any limitations on this connection.

[0106] For ease of understanding, please refer to the following: Figure 2 , Figure 2 This is a schematic diagram illustrating a scenario for pushing media data, as provided in an embodiment of this application. Wherein, as... Figure 2 The user terminal A shown can be in the above Figure 1 Any user terminal selected in the user terminal cluster of the corresponding embodiment, for example, the user terminal can be the aforementioned user terminal 100a; such as Figure 2 The service server shown can be one of the above-mentioned... Figure 1 The corresponding embodiment is the business server 1000.

[0107] It should be understood that, such as Figure 2 The scenario shown is an example where the first media type is video and the second media type is text. Figure 2 As shown, when user a uses user terminal A, they can view public broadcast media data (which may include video media data and text media data) pushed to user a through the information push application installed on user terminal A. Figure 2As shown, at 8:10 AM on March 16, 2021, the public broadcast media data pushed to the target user included media data 20a, media data 20b, media data 20c, and media data 20d. This public broadcast media data, including media data 20a, 20b, 20c, and 20d, can be referred to as the first public broadcast media data. Media data 20a was presented in the form of an image and text, and its data summary was "A man in City A broke down in tears after returning to his hometown after the Spring Festival." Media data 20b was also presented in the form of an image and text, and its data summary was... The first public broadcast media data includes "High school teachers and students singing, a girl's voice instantly captivates the audience"; Media data 20c is presented in the form of images and text, and the data summary of Media data 20d is "College students show off their peeling tool, no injury to hands, no knife needed"; it should be understood that Media data 20a, Media data 20b, and Media data 20c are all text-based media data, and all belong to the text type; Media data 20d is presented in the form of video and text, and the data summary of Media data 20d is "Man encounters tiger while chopping wood," and Media data 20d is video media data, belonging to the video type. For this first public broadcast media data, which includes both video and text types, user A can click on any media data to browse and view it. After user A clicks on any media data, they can view the complete content of that media data. For example, after user A clicks on Media data 20a, they can view all the images and the complete text content of Media data 20a.

[0108] It should be noted that if the video media data in the public broadcast media data pushed to the target user has the auto-play attribute, then the target user does not need to click on the media data; the corresponding video content of the video media data will play automatically. For example, if the playback attribute of media data 20d is auto-play, then user a does not need to click on media data 20d to play its video content.

[0109] It should be understood that the media data push time for the first public broadcast media data was 8:10 AM on March 16, 2021. The next media data push time for user a (e.g., the second public broadcast media data) can be obtained (e.g., 3:10 PM on March 16, 2021). The media data push time of 8:10 AM on March 16, 2021 can be used as the starting time, and the push time can be between the two media data push times (i.e., 8:10 AM to 3:10 PM on March 16, 2021). Between 8:10 AM and 3:10 PM on March 16, 2021, select a time as the end time (e.g., select 3:00 PM on March 16, 2021 as the end time); then, the time period consisting of the start time (8:10 AM on March 16, 2021) and the end time (3:00 PM on March 16, 2021) can be used as the first time period; within the first time period, user environment information corresponding to user a, as well as the browsing behavior association information of user a for the first public broadcast media data can be obtained. The user environment information may include user attribute information and environment attribute information. The user attribute information may be the attribute information of user A, such as user A's age, gender, and geographic location. The environment attribute information may include the media data push time (e.g., 8:10 AM on March 16, 2021), refresh count, version information of the information push application, operating system information of user terminal A, network status information, playback attribute information of the first public broadcast media data (e.g., whether it autoplays), and launch source information, etc. The browsing behavior association information may be the click behavior association information of user A regarding the first public broadcast media data, such as the total number of pushes of the first public broadcast media data (e.g., ...). Figure 2 As shown, there are 4), and the number of video-type media data pushed in the first public broadcast media data (e.g. Figure 2 As shown, (1), the total number of clicks by user a on the first public broadcast media data (e.g., the total number of clicks is 10), and the number of clicks by user a on the video-type media data in the first public broadcast media data (e.g., the number of clicks is 5).

[0110] Furthermore, user terminal A can send user environment information and browsing behavior association information of user a to the business server. The business server can generate user browsing status features corresponding to user a based on the user environment information and browsing behavior association information. Subsequently, the business server can predict a first predicted quantity (e.g., the first predicted quantity is 2) based on the user browsing status features. This first predicted quantity is the number of video-type media data included in the second public broadcast media data pushed to user a at 15:10 on March 16, 2021, the media data push time. The quantity and configuration total quantity can determine the second predicted quantity (i.e., the quantity of text-type media data included in the second public broadcast media data pushed to user a at 15:10 on March 16, 2021, the media data push time); wherein, the configuration total quantity can be the preset total quantity of media data required to be pushed to user a, and the total quantity of the first public broadcast media data and the second public broadcast media data can both be the configuration total quantity, which can be a manually specified value, such as, if the configuration total quantity is 4, then the second predicted quantity can be 2 (the result obtained by subtracting the first predicted quantity 2 from the configuration total quantity 4).

[0111] Furthermore, the business server can retrieve two media data entries belonging to the video type from the database (e.g., such as...). Figure 2 The media data 20g and media data 20h shown are used to obtain two text-type media data (e.g., such as...). Figure 2 The media data 20e and 20f shown are included. It should be understood that the public broadcast media data 20, composed of media data 20g, 20h, 20e, and 20f, can be referred to as the second public broadcast media data. It should be understood that the service server can return the public broadcast media data 20 to user terminal A, and user terminal A can push the public broadcast media data 20 to user A at 15:10 on March 16, 2021, the media data push time. Figure 2As shown, at 15:10 on March 16, 2021, when the media data push time arrives, user a can view the public broadcast media data 20 (including media data 20e, media data 20f, media data 20g and media data 20h) on the display interface of user terminal A. Among them, media data 20e is presented in the form of an image and text, with the data summary "Heavy snowfall in City B, citizens capture beautiful scenes"; media data 20f is also presented in the form of an image and text, with the data summary "Woman's identity information on a plane is abnormal, the result is unexpected"; it should be understood that media data 20e and 20f are both text-based media data, belonging to the text type; media data 20g is presented in the form of a video and text, with the data summary "Sandstorm hits City A, giant sand wall rushes in"; media data 20h is presented in the form of a video and text, with the summary "Girl's dance in heavy snow amazes everyone"; it should be understood that media data 20g and 20h are both video media data, belonging to the video type. User A can click on any media data in the public broadcast media data 20 to browse and view.

[0112] Understandably, for the first public broadcast media data, user A's total click count is 10, while for the video-type media data within that first public broadcast media data, user A's click count is 5. Since the number of video-type media data in the first public broadcast media data is only 1, and the number of video-type media data is extremely small, user A's click count for video-type media data within the first public broadcast media data still accounts for a relatively high proportion of the total click count. This indicates that user A currently prefers to browse video-type media data. Therefore, the first predicted number based on user A's current state characteristics is greater than the number of video-type media data in the first public broadcast media data. In other words, the number of video-type media data in the second public broadcast media data has been appropriately increased. This means that, with the total number of data types remaining unchanged, the ratio of video-type to text-type media data pushed to user A in the second public broadcast media data is more in line with user A's preferences, thus improving the accuracy of media data type recommendation.

[0113] Further, please see Figure 3 , Figure 3 This is a flowchart illustrating a data processing method provided in an embodiment of this application. The method can be implemented by a user terminal (e.g., the one described above). Figure 1The operation can be performed by any user terminal in the user terminal cluster shown, such as user terminal 100b, or by a service server (e.g., the one mentioned above). Figure 1 The method can be executed by the business server 1000 shown, or it can be jointly executed by the user terminal and the business server. For ease of understanding, this embodiment will use the execution of the method by the aforementioned business server as an example.

[0114] The method may include at least the following steps S101-S104:

[0115] Step S101: Obtain user environment information corresponding to the target user and related information on the target user's browsing behavior of the first public broadcast media data within the first time period.

[0116] In this application, the first public broadcast media data can be media data pushed by the service server to the target user terminal (the terminal corresponding to the target user). This first public broadcast media data may include media data of different types. For example, it may include media data belonging to a first media type and media data belonging to a second media type. It should be understood that when the target user uses the target application (e.g., an information push application) on the target user terminal, the target user can view the first public broadcast media data within that application (which is installed on the target user terminal). The target user can click on any media data under any media type within the first public broadcast media data to browse and view it. Both the first and second media types can include video, text, music, etc. When the first media type is a certain media type, the second media type is any other media type besides that type. For example, when the first media type is video, the second media type is any other media type besides video, such as music, text, etc.

[0117] The first time period in this application can be a time period that starts at the first media data push time and ends earlier than the second media data push time. The first media data push time can refer to the time when the first public broadcast media data is pushed to the target user terminal; the second media data push time can refer to the time after the first public broadcast media data is pushed to the target user terminal for the next time public broadcast media data (such as the second public broadcast media data) is pushed. In other words, the first time period can be a time period between two consecutive media data push times. For example, if the first public broadcast media data is pushed to the target user terminal at 14:00 on March 16, 2021, and the next time public broadcast media data is pushed to the target user terminal is at 15:00 on March 16, 2021, then the push time of 14:00 on March 16, 2021 can be used as the start time. Between 14:00 on March 16, 2021 and 15:00 on March 16, 2021, a time can be selected as the end time (for example, 14:50 on March 16, 2021). Thus, the time period with the start time of 14:00 on March 16, 2021 and the end time of 14:50 on March 16, 2021 can be obtained, and this time period can be the first time period. It should be understood that the timing of each media data push can be artificially determined. For example, the first push of public broadcast media data (i.e., the initial media data push time) can be artificially determined to be 08:00 on March 16, 2021, and then a public broadcast media data push can be performed at fixed intervals. This determines the timing of each media data push. For example, if the fixed interval is 4 hours, then public broadcast media data will be pushed every 4 hours. Since the first public broadcast media data push was at 08:00 on March 16, 2021, and the interval is 4 hours, then the second public broadcast media data push... The first public broadcast of media data was at 12:00 on March 16, 2021, and the second public broadcast of media data was at 16:00 on March 16, 2021, and so on. It should be understood that the time for each media data push can also be randomly selected. For example, if the first public broadcast of media data was at 08:00 on March 16, 2021, a random time can be selected after 08:00 on March 16, 2021 as the time for the second public broadcast of media data. For example, 20:00 on March 16, 2021 can be randomly selected as the time for the second public broadcast of media data.

[0118] It should be understood that after the first public broadcast media data is pushed to the target user's terminal, the target user can click to view the first public broadcast media data and obtain the target user's user environment information within the first time period and the browsing behavior related information (i.e., click behavior related information) related to the first public broadcast media data. The user environment information and the browsing behavior related information can together form the target user's user status information within the first time period. The user environment information may include user attribute information and environment attribute information. The user attribute information may include the target user's age, gender, and geographic location. The environment attribute information may include the media data push time (e.g., 14:00 on March 16, 2021), refresh count, version information of the target application, operating system information of the target user's terminal, network status information, playback attribute information of the first public broadcast media data (e.g., whether it plays automatically), and launch source information. The browsing behavior association information may include the total number of pushes of the first public broadcast media data (e.g., 10), the number of pushes of media data belonging to the first media type within the first public broadcast media data (e.g., 6), the total number of clicks by the target user on the first public broadcast media data (e.g., 20 total clicks), and the number of clicks by the target user on media data belonging to the first media type within the first public broadcast media data (e.g., 10 clicks).

[0119] It is understandable that the time during which the user environment information corresponding to the target user and the browsing behavior information of the target user regarding the first public broadcast media data are obtained within the first time period can be considered as the end time of the first time period. For example, if the first time period starts at 14:00 on March 16, 2021 and ends at 14:50 on March 16, 2021, then the time during which the user environment information corresponding to the target user and the browsing behavior information of the target user regarding the first public broadcast media data are obtained can be considered as the end time of 14:50 on March 16, 2021.

[0120] Step S102: Based on user environment information and browsing behavior-related information, determine the user browsing status characteristics of the target user for the first public broadcast media data.

[0121] In this application, based on the user's environmental information (including user attribute information and environmental attribute information) and browsing behavior association information, the user status information of the target user within a first time period can be constructed. Based on this user status information, the user browsing status characteristics of the target user regarding the first public broadcast media data within the first time period can be determined. The specific determination method can be found in the following sections. Figure 5 The description in the corresponding embodiments.

[0122] Step S103: Predict the first predicted quantity corresponding to the target user based on the user's browsing status characteristics; the first predicted quantity refers to the number of media data belonging to the first media type pushed to the terminal corresponding to the target user within the second time period; the first time period is earlier than the second time period.

[0123] In this application, the first predicted quantity corresponding to the target user can be predicted by using the target quantity prediction model and the aforementioned user browsing status characteristics. The specific method is as follows: the user browsing status characteristics can be input into the target quantity prediction model, and the target quantity prediction model can output the prediction probabilities corresponding to N configuration quantities respectively, thus obtaining N prediction probabilities; wherein, each configuration quantity represents the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user in the second time period; N is an integer; the maximum prediction probability is obtained from the N prediction probabilities, and the configuration quantity corresponding to the maximum prediction probability is determined as the first predicted quantity.

[0124] It is understandable that the second time period can be a period starting with the second media data push time and ending earlier than the third media data push time. The second media data push time can refer to the time when the second public broadcast media data is pushed to the target user terminal; the third media data push time can refer to the time after the push of the aforementioned second public broadcast media data, the next time public broadcast media data (such as third public broadcast media data) is pushed to the target user terminal. Within the second time period, the second public broadcast media data pushed to the target user's terminal may include media data belonging to the first media type and media data belonging to the second media type. The number of media data belonging to the first media type within the second public broadcast media data (i.e., the first predicted number) can be determined through the user's browsing status characteristics.

[0125] It is understandable that the N configuration quantities can refer to the number of media data configured for the first media type. For example, if the first media type is video, then the N configuration quantities can be the N quantities configured for media data belonging to the video type, i.e., N configuration video quantities. When the first media type is video, the first prediction quantity can be the predicted video quantity. The N configuration quantities can be manually defined values, and can be integers from 0 to N. For example, N can be manually defined as 20, then the N configuration quantities can be 0, 1, 2, ... 20.

[0126] It is understandable that the prediction probabilities corresponding to the N configuration quantities output by the target quantity prediction model can refer to the quality assessment scores corresponding to the N configuration quantities output by the target quantity prediction model under the user's browsing state characteristics. The higher the prediction probability (i.e. the higher the quality assessment score), the better the corresponding configuration quantity matches the user's browsing state characteristics.

[0127] Step S104: Determine the second predicted quantity based on the first predicted quantity; the second predicted quantity refers to the quantity of media data belonging to the second media type pushed to the terminal corresponding to the target user within the second time period.

[0128] In this application, based on the user browsing status characteristics of the target user regarding the first public broadcast media data in the first time period, the number of media data belonging to the first media type in the second public broadcast media data pushed to the target user in the second time period (i.e., the first predicted number) can be predicted. Based on this first predicted number, the number of media data belonging to the second media type in the second public broadcast media data pushed to the target user in the second time period (i.e., the second predicted number) can be determined. Specifically, the method can be as follows: A total configured quantity can be obtained; wherein, the total configured quantity refers to the total number of media data required to be pushed to the terminal corresponding to the target user; the total configured quantity can be subtracted from the first predicted quantity to obtain the second predicted number.

[0129] It should be understood that the total configured quantity is the total amount of media data required to be pushed to the terminal corresponding to the target user each time. This total quantity can remain unchanged; that is, the total amount of the first public broadcast media data and the second public broadcast media data are the same. This total configured quantity can be a manually specified value. Optionally, the total amount of media data required to be pushed to the terminal corresponding to the target user each time can also be a different value (that is, the total amount of the first public broadcast media data and the second public broadcast media data are different). A value can be randomly selected within a certain range (e.g., within 0-20) as the total amount of public broadcast media data pushed each time. For example, the value 10 can be selected as the total amount of the first public broadcast media data, and the value 9 can be selected as the total amount of the second public broadcast media data. Then, after determining the first predicted quantity, the total quantity specified when pushing the second public broadcast media data (i.e., the total configured quantity) can be obtained. Subtracting the first predicted quantity from the total configured quantity yields the second predicted quantity.

[0130] Step S105: Based on the first predicted quantity and the second predicted quantity, push the second public broadcast media data to the terminal corresponding to the target user within the second time period; the second public broadcast media data includes first target media data belonging to the first media type and second target media data belonging to the second media type; the quantity of the first target media data is the first predicted quantity; the quantity of the second target media data is the second predicted quantity.

[0131] In this application, the second public broadcast media data can be pushed to the terminal corresponding to the target user (i.e., the target user terminal) when the second media data push time is reached (which can be the start time of the second time period). The specific method is as follows: a set of media data to be pushed can be obtained from a database; wherein the set of media data to be pushed includes media data of a first media type expected to be pushed to the terminal corresponding to the target user within the second time period, and media data of a second media type expected to be pushed to the terminal corresponding to the target user; based on a first predicted quantity, first target media data can be obtained from the set of media data to be pushed; the first target media data belongs to the first media type, and the quantity of the first target media data is the first predicted quantity; based on a second predicted quantity, second target media data can be obtained from the set of media data to be pushed; the second target media data belongs to the second media type, and the quantity of the second target media data is the second predicted quantity; subsequently, the first target media data and the second target media data can be determined as the second public broadcast media data, and the second public broadcast media data can be pushed to the terminal corresponding to the target user within the second time period (i.e., at the second media data push time).

[0132] It should be understood that when predicting the initial quantity of a target user based on user browsing status characteristics, a target quantity prediction model can be used. To make the prediction results of the target quantity prediction model more accurate, the quantity prediction model can be trained first to obtain the target quantity prediction model, and then the target quantity prediction model can be applied. For ease of understanding, please refer to [further details omitted]. Figure 4 , Figure 4 This is an architecture diagram provided in an embodiment of this application for training and applying a model. For example... Figure 4 As shown, the architecture may include a sample generation module 40a, a feature acquisition module 40b, a target quantity prediction model 40c, a quantity selection module 40d, and a data storage module 40e; all of these modules can be deployed on a business server. For ease of understanding, the following will describe each module in the architecture:

[0133] The sample generation module 40a can be used to generate training sample data for training the quantity prediction model based on the data sent by the business server. It should be understood that the business server can train and update the quantity prediction model based on the training sample data generated by the sample generation module 40a, thereby obtaining the target quantity prediction model 40c. The specific implementation process for generating training sample data and training and updating the quantity prediction model based on the training sample data can be found in subsequent sections. Figure 6 The description in the corresponding embodiments.

[0134] The feature acquisition module 40b is used to acquire user environment information and browsing behavior association information of the target user, and determine the user browsing state characteristics of the target user based on the user environment information and browsing behavior association information. For the specific implementation method of determining the user browsing state characteristics of the target user based on the user environment information and browsing behavior association information, please refer to the above. Figure 3 The description of the user browsing status characteristics of the target user in step S102 of the corresponding embodiment will not be repeated here.

[0135] The target quantity prediction model 40c can input the user browsing status features determined by the feature acquisition module 40b. Through the target quantity prediction model 40c and the user browsing status features, the prediction probabilities corresponding to the N configuration quantities can be output. The prediction probabilities here can refer to the quality evaluation scores. That is to say, through the target quantity prediction model 40c and the user browsing status features, the N configuration quantities can be scored respectively, thereby obtaining the quality evaluation scores corresponding to the N configuration quantities.

[0136] The quantity selection module 40d can be used to select the first predicted quantity corresponding to the user browsing state feature from N configuration quantities. It should be understood that the quality evaluation scores corresponding to the N configuration quantities output by the above target quantity prediction model 40c can be input into the quantity selection module 40d. In the quantity selection module 40d, the maximum quality evaluation score among the N quality evaluation scores can be obtained, and the configuration quantity corresponding to the maximum quality evaluation score can be determined as the first predicted quantity corresponding to the user browsing state feature.

[0137] The data storage module 40e can be used to store data. The quantity selection module 40d can send data (e.g., a first predicted quantity) to the data storage module 40e, which can then store the first predicted quantity. It should be understood that the business server can retrieve data from the data storage module 40e.

[0138] In this embodiment, user environment information corresponding to a target user within a first time period and browsing behavior association information of the target user regarding the first public broadcast media data can be obtained. Based on the user environment information and browsing behavior association information, the user browsing status characteristics of the target user (i.e., the target user's current preference status characteristics for different types of media data) can be determined. Based on the user browsing status characteristics, a first predicted quantity corresponding to the target user can be predicted. Based on the first predicted quantity, a second predicted quantity can be determined. Based on the first and second predicted quantities, the second public broadcast media data can be pushed to the target user within the second time period. The second public broadcast media data includes media data belonging to a first media type (i.e., the first target media data) and media data belonging to a second media type (i.e., the second target media data). The quantity of the first target media data is the first predicted quantity, and the quantity of the second target media data is the second predicted quantity. In other words, this application can predict, based on the user browsing status characteristics of the target user in the first time period, the number of media data belonging to the first media type and the number of media data belonging to the second media type when pushing second public broadcast media data to the target user in the second time period. Since the number of media data belonging to the first media type and the number of media data belonging to the second media type are predicted and determined based on the user browsing status characteristics (current preference status characteristics) of the target user, the number of media data belonging to the first media type and the number of media data belonging to the second media type are consistent with the target user's current preference status for the first and second media types, resulting in high accuracy. In summary, this application can improve the accuracy of media data type recommendation.

[0139] To better understand the specific process of pushing media data to target users, please refer to [link / reference]. Figure 5 , Figure 5 This is a schematic diagram illustrating the push of media data provided in an embodiment of this application. For example... Figure 5 The user terminal A shown can be in the above Figure 1 Any user terminal selected in the user terminal cluster of the corresponding embodiment, for example, the user terminal can be the aforementioned user terminal 100a; such as Figure 5 The service server shown can be one of the above-mentioned... Figure 1 The corresponding embodiment is the business server 1000.

[0140] like Figure 5As shown, user 'a' can be the target user. User terminal A can obtain the target user's user environment information (including user attribute information and environment attribute information) and the target user's browsing behavior association information regarding the first public broadcast media data within the first time period. User terminal A can send the user environment information and browsing behavior association information to the business server. The business server can determine the target user's browsing status characteristics regarding the first public broadcast media data based on the user environment information and browsing behavior association information. The specific method can be as follows: a first hash value corresponding to the user attribute information can be generated based on the user attribute information and a hash function; a second hash value corresponding to the environment attribute information can be generated based on the environment attribute information and a hash function; a third hash value corresponding to the browsing behavior association information can be generated based on the browsing behavior association information and a hash function; subsequently, the target user's browsing status characteristics regarding the first public broadcast media data can be determined based on the first hash value, the second hash value, and the third hash value.

[0141] The specific method for determining the user browsing status characteristics of the target user for the first public broadcast media data based on the first hash value, the second hash value, and the third hash value can be as follows: A vector mapping table can be obtained; wherein the vector mapping table includes the mapping relationship between configuration hash values ​​and configuration vectors; the configuration hash values ​​include the first hash value, the second hash value, and the third hash value; subsequently, the configuration vector that has a mapping relationship with the first hash value can be obtained from the vector mapping table as the first user feature vector; the configuration vector that has a mapping relationship with the second hash value can be obtained from the vector mapping table as the second user feature vector; the configuration vector that has a mapping relationship with the third hash value can be obtained from the vector mapping table as the third user feature vector; subsequently, the first user feature vector, the second user feature vector, and the third user feature vector can be fused to obtain the user browsing status characteristics of the target user for the first public broadcast media data.

[0142] It should be understood that the acquired user attribute information, environmental attribute information, and browsing behavior status information can all be represented by strings. These strings can be hashed using a hash function to generate hash values. Each hash value corresponds to a vector, which can be used as the feature vector for that user attribute information, environmental attribute information, or browsing behavior status information. For example, taking age as an example, if the target user's age is 18, this age can be represented by the string "age_18". Hash calculation of this string "age_18" using a hash function generates a hash value A1, which can be called the first hash value. The vector corresponding to hash value A1 is vector B1, which can be used as the user feature vector for that user attribute information, and vector B1 can be called the first user feature vector. Understandably, if user attribute information, environment attribute information, and browsing behavior status information are each composed of multiple data information, then the string corresponding to each data information in the user attribute information, environment attribute information, or browsing behavior status information can be obtained, and the hash value corresponding to each string can be calculated by a hash function, thereby determining the feature vector corresponding to each data information. Subsequently, the feature vectors corresponding to each data information can be fused (for example, by adding or concatenating) to obtain the user feature vectors corresponding to the user attribute information, environment attribute information, or browsing behavior status information respectively. For example, taking user attribute information including age and gender as an example, both age and gender can be referred to as data information. The target user's age is 18 years old and the gender is male. The string corresponding to the age can be "age_18" and the string corresponding to the gender can be "sex_man". By performing hash calculations on the strings "age_18" and "sex_man" respectively, we can determine that the hash value corresponding to the string "age_18" is hash value A1 and the hash value corresponding to the string "sex_man" is hash value A2. We can obtain the vector corresponding to hash value A1 as vector B1 and the vector corresponding to hash value A2 as vector 2. Then, vector B1 can be used as the feature vector of the age and vector B2 can be used as the feature vector of the gender. Subsequently, vector B1 and vector B2 can be added together, and the result can be used as the user feature vector corresponding to the user attribute information.

[0143] It should be understood that after determining the first user feature vector corresponding to user attribute information, the second user feature vector corresponding to environmental attribute information, and the third user feature vector corresponding to browsing behavior association information, the first user feature vector, the second user feature vector, and the third user feature vector can be fused. The fusion method can be concatenation, addition, etc. The feature vector obtained after fusion can be determined as the user browsing status feature of the target user.

[0144] Furthermore, the user's browsing status features can be input into the target quantity prediction model, which can then output a first predicted quantity corresponding to the user's browsing status features. For the specific implementation of the target quantity prediction model in determining the first predicted quantity, please refer to the above. Figure 3 The description of determining the first predicted quantity in step S103 of the corresponding embodiment will not be repeated here. Further, the total configuration quantity can be obtained, and the second predicted quantity can be determined based on the total configuration quantity and the first predicted quantity. For the specific implementation of determining the second predicted quantity, please refer to the above. Figure 3 The description of determining the second predicted quantity in step S104 of the corresponding embodiment will not be repeated here. Further, based on the first and second predicted quantities, the second public broadcast media data can be determined, and this second public broadcast media data can be returned to user terminal A. User terminal A can then display the second public broadcast media data on its display interface, allowing user a to view it. For the specific implementation of determining the second public broadcast media data, please refer to the above. Figure 3 The description of step S105 in the corresponding embodiment will not be repeated here.

[0145] Further, please see Figure 6 , Figure 6 This is a flowchart illustrating another data processing method provided in an embodiment of this application. This data processing method can be a model training method, such as... Figure 6 As shown, the process may include at least the following steps S201-S203:

[0146] Step S201: Obtain the browsing status features of a first sample user and a second sample user associated with the target user; the browsing status features of the first sample user are determined based on the target user's first sample user environment information obtained within the first sample time period, and the target user's first browsing behavior related information regarding the first sample public broadcast media data; the browsing status features of the second sample user are determined based on the target user's second sample user environment information obtained within the second sample time period, and the target user's second browsing behavior related information regarding the second sample public broadcast media data; the first sample time period is earlier than the second sample time period.

[0147] In this application, the specific method for obtaining the browsing status characteristics of the first sample user and the browsing status characteristics of the second sample user can be as follows: Within the first sample time period, the environmental information of the first sample user corresponding to the target user, and the association information of the target user's first sample browsing behavior towards the first sample public broadcast media data, can be obtained; the browsing status characteristics of the target user can be determined based on the environmental information of the first sample user and the association information of the first sample browsing behavior; subsequently, a target configuration quantity can be selected from N configuration quantities as the number of first sample types corresponding to the browsing status characteristics of the first sample user; wherein, the number of first sample types can refer to the number of media data belonging to the first media type pushed to the terminal corresponding to the target user within the second sample time period; subsequently, the total configuration quantity can be obtained; the total configuration quantity refers to the total number of media data required to be pushed to the terminal corresponding to the target user; based on the total configuration quantity and the number of first sample types, the browsing status characteristics of the second sample user can be determined.

[0148] The specific method for determining the browsing status characteristics of the second sample user based on the total number of configurations and the number of the first sample types can be as follows: The total number of configurations can be subtracted from the number of the first sample types to obtain the number of second sample types of media data belonging to the second media type pushed to the terminal corresponding to the target user during the second sample time period. Based on the number of the first sample types and the number of the second sample types, the second sample public broadcast media data can be pushed to the terminal corresponding to the target user during the second sample time period. The second sample public broadcast media data includes first sample media data belonging to the first media type and second sample media data belonging to the second media type. The number of first sample media data is equal to the number of the first sample types, and the number of second sample media data is equal to the number of the second sample types. Subsequently, during the second sample time period, the second sample user environment information corresponding to the target user and the second sample browsing behavior association information of the target user regarding the second sample public broadcast media data can be obtained. Based on the second sample user environment information and the second sample browsing behavior association information, the second sample user status characteristics of the target user can be determined.

[0149] It should be understood that at a certain time (e.g., 20:10 on March 16, 2021), the first sample of public broadcast media data can be pushed to the target user's terminal. The target user can browse and view the first sample of public broadcast media data through the target application on the target user's terminal. It should be understood that the time when the first sample of public broadcast media data is pushed can be referred to as the first sample media data push time, and the time when the next public broadcast media data (e.g., the second sample of public broadcast media data) is pushed after the first sample of public broadcast media data is pushed can be referred to as the second sample media data push time. After the first sample of public broadcast media data is pushed to the target user's terminal at the first sample media data push time, the first sample user environment information corresponding to the target user and the first sample browsing behavior association information related to the first public broadcast media data can be obtained. A time can be selected between the first sample media data push time and the second sample media data push time as the end time for obtaining the target user's information, thereby obtaining the first sample time period from the start time (the first sample media data push time) to the end time. In other words, within the first sample time period, the target user's first sample user environment information and the browsing behavior related information of the first public broadcast media data can be obtained.

[0150] Furthermore, the browsing state characteristics of the target user can be determined based on the association information between the first sample user's environment information and the first sample browsing behavior. For the specific implementation of determining the browsing state characteristics of the target user, please refer to the above. Figure 3 The description of determining the user browsing status characteristics of the target user in step S102 of the corresponding embodiment will not be repeated here. Subsequently, a certain configuration quantity (e.g., 5) can be randomly selected from N configuration quantities (e.g., when N is 21, the N configuration quantities may include 0, 1, ..., 21) as the target configuration quantity, and this target configuration quantity is used as the first sample type quantity corresponding to the first sample user browsing status characteristics. The first sample type quantity can be the number of media data belonging to the first media type included in the second sample public broadcast media data when the next sample public broadcast media data (e.g., the second sample public broadcast media data) push time (i.e., the second sample media data push time) arrives. It should be understood that N configuration quantities can refer to the number of media data belonging to the first media type configured. When the first media type is a music type, the N configuration quantities can be N configured music quantities, and the first sample type quantity can be the sample music quantity.

[0151] It should be understood that the total number of public broadcast media data pushed to the target user can be referred to as the total configuration quantity. This total configuration quantity can be obtained, and by subtracting the first sample type quantity from it, the second sample type quantity can be obtained (the quantity of media data belonging to the second media type included in the second sample public broadcast media data pushed to the target user at the second sample media data push time). Furthermore, the database can retrieve the first sample media data (sample media data belonging to the first media type) in quantity equal to the first sample type quantity, and the second sample media data (sample media data belonging to the second media type) in quantity equal to the second sample type quantity. The media data composed of the first sample media data and the second sample media data can be referred to as the second sample public broadcast media data.

[0152] Furthermore, when the second sample media data push time arrives, the second sample public broadcast media data can be pushed to the target user's terminal, and the target user can browse and view the second sample public broadcast media data. Similarly, it should be understood that after the second sample public broadcast media data is pushed, the time of the next push of sample public broadcast media data (e.g., the third sample public broadcast media data) can be referred to as the third sample media data push time. After the second sample media data is pushed to the target user's terminal when the second sample media data push time arrives, the second sample user environment information corresponding to the target user, as well as the second sample browsing behavior association information related to the second public broadcast media data, can be obtained. A time can be selected between the second media data push time and the third media data push time as the end time for obtaining the target user's information, thus obtaining the second sample time period from the start time (the second sample media data push time) to the end time. That is to say, within the second sample time period, the second sample user environment information of the target user and the second sample browsing behavior association information related to the second public broadcast media data can be obtained. Based on the second sample user environment information and the second sample browsing behavior association information, the second sample user browsing status characteristics of the target user can be determined. For the specific implementation method of determining the second sample user browsing status characteristics of the target user, please refer to the above. Figure 3 The description of the user browsing status characteristics of the target user in step S102 of the corresponding embodiment will not be repeated here.

[0153] Step S202: Based on the browsing status characteristics of the first sample user and the browsing status characteristics of the second sample user, determine the training sample data associated with the target user.

[0154] In this application, the browsing behavior association information may include the total number of pushes of the public broadcast media data, the number of pushes of the public broadcast media data belonging to the first media type, the total number of clicks by the target user on the public broadcast media data, and the number of clicks by the target user on the media data belonging to the first media type in the public broadcast media data. Therefore, the first sample browsing behavior association information and the second sample browsing behavior association information may also include the total number of pushes of the sample public broadcast media data, the number of pushes of the sample media data belonging to the first media type in the sample public broadcast media data, the total number of clicks by the target user on the sample public broadcast media data, and the number of clicks by the target user on the sample media data belonging to the first media type in the sample public broadcast media data. The total number of clicks by target users on sample public broadcast media data can be referred to as the sample click frequency. Based on the browsing status characteristics of the first sample user, the browsing status characteristics of the second sample user, and the sample click frequency, training sample data for training the quantity prediction model can be determined. Specifically, the sample click frequency can be determined as a feedback parameter associated with the target user and the number of the first sample type. Subsequently, the data set consisting of the browsing status characteristics of the first sample user, the number of the first sample type, the feedback parameter, and the browsing status characteristics of the second sample user can be determined as the training sample data associated with the target user.

[0155] For example, taking a video as the first media type and a text as the second media type, the first sample user's browsing state feature, determined by associating the first sample user's environment information with the first sample browsing behavior information, is s1. Among N configured video quantities (e.g., including 0, 1, ..., 21), the number of sample videos selected for the first sample user's browsing state feature s1 (i.e., the number of first sample types) is 6. Since the total number of configurations is 8, the number of sample texts (i.e., the number of second sample types) can be 2 (the total number of configurations is 8 minus the number of sample videos 6). Subsequently, the sample video media data with the number of sample videos and the sample text data with the number of sample texts can be... This media data is identified as the second sample public broadcast media data. This second sample public broadcast media data can be pushed to the target user's terminal. The target user can browse and view the sample video media data and sample text media data. At this time, the target user's second sample user environment information and second sample browsing behavior correlation information regarding the second sample public broadcast media data can be obtained. The sample click frequency r1 in the second sample browsing behavior correlation information can be used as a feedback parameter corresponding to the number of sample videos. This feedback parameter can be a parameter used to evaluate the quality of the number of sample videos; a higher sample click frequency r1 indicates a better number of sample videos. It should be understood that based on the second sample user environment information and the second sample browsing behavior correlation information, the second sample user browsing state feature s2 corresponding to the target user can be determined. The first sample user browsing state feature s1, the number of sample videos 6, the feedback parameter (i.e., the sample click frequency r1), and the second sample user browsing state feature s2 can form a data set (s2, 6, r1, s2). This data set (s2, 6, r1, s2) can then be used as training sample data for training the quantity prediction model.

[0156] It should be noted that the data set composed of the first sample user browsing state features, the number of first sample types, feedback parameters, and the second sample user browsing state features can determine the training sample data associated with the target user (e.g., the aforementioned data set (s2, 6, r1, s2)). This training sample data can be used as one of the training sample data for training the quantity prediction model, and can be referred to as the first training sample data. It should be understood that the second sample user browsing state features, the number of sample types corresponding to the second sample user browsing state features, the feedback parameters corresponding to the second sample user browsing state features, and the third sample user browsing state features can determine the second training sample data used to train the quantity prediction model. In other words, by obtaining the sample user browsing state features twice (e.g., the first sample user browsing state features and the second sample user browsing state features), one training sample data for training the quantity prediction model can be determined. Therefore, by obtaining the sample user browsing state features twice, multiple training sample data for training the quantity prediction model can be determined.

[0157] Step S203: Train the quantity prediction model based on the training sample data to obtain the target quantity prediction model; the target quantity prediction model is used to predict the number of types corresponding to the browsing status characteristics of the target user, and the number of types refers to the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user.

[0158] In this application, the specific method for training a quantity prediction model based on training sample data to obtain a target quantity prediction model can be as follows: The first sample user browsing state feature from the training sample data can be input into the quantity prediction model. Through the quantity prediction model and the first sample user browsing state feature, N first sample prediction probabilities corresponding to the number of configured videos can be output, resulting in N first sample prediction probabilities. Among these N first sample prediction probabilities, the first sample prediction probability corresponding to the number of first sample types in the training sample data can be determined and used as the first probability to be calculated. Subsequently, the second sample user browsing state feature from the training sample data can be input into the quantity prediction model. Through the quantity prediction model and the second sample user browsing state feature, N first sample prediction probabilities can be output. The state features can output the second sample prediction probabilities corresponding to N configuration quantities, resulting in N second sample prediction probabilities. Among the N second sample prediction probabilities, the maximum second sample prediction probability can be obtained, and the configuration quantity corresponding to the maximum second sample prediction probability can be determined as the optimal type quantity. Subsequently, the second sample user browsing state features can be input into the quantity evaluation model. Through the quantity evaluation model and the second sample user browsing state features, the third sample prediction probabilities corresponding to N configuration quantities can be output, resulting in N third sample prediction probabilities. Based on the first probability to be calculated, the optimal type quantity, the feedback parameters in the training sample data, and the N third sample prediction probabilities, the quantity prediction model can be trained to obtain the target quantity prediction model.

[0159] The specific method for training the quantity prediction model based on the first probability to be calculated, the optimal number of types, the feedback parameters in the training sample data, and the prediction probabilities of N third samples to obtain the target quantity prediction model can be as follows: From the N prediction probabilities of the third samples, obtain the third sample prediction probability corresponding to the optimal number of types and use it as the second probability to be calculated; obtain the discount coefficient of the quantity prediction model, and based on the discount coefficient, the feedback parameters, and the second probability to be calculated, determine the target calculation probability associated with the training sample data; subtract the target calculation probability from the first probability to be calculated to obtain the calculation result, and determine the loss function value of the quantity prediction model based on the calculation result; train the quantity prediction model based on the loss function value to obtain the target quantity prediction model. It should be understood that the optimal number of types can correspond to the first media type; for example, if the first media type is video, then the optimal number of types can be the optimal number of videos.

[0160] Understandably, the first probability to be computed is obtained by the quantitative prediction model predicting the browsing state features of the first sample user in the training sample data. This first probability to be computed can represent the value (i.e., the sample prediction probability) of the number of first sample types corresponding to the browsing state features of the first sample user, output by the quantitative prediction model. This value can be used to characterize the quality score of the number of first sample types corresponding to the browsing state features of the first sample user. The second probability to be computed is obtained by the quantitative prediction model and the quantitative evaluation model jointly predicting the browsing state features of the first sample user in the training sample data (i.e., the quantitative prediction model first predicts the optimal number of types based on the browsing state features of the first sample user, and then the quantitative evaluation model predicts the optimal number of types based on the browsing state features of the first sample user, obtaining the sample prediction probability corresponding to the optimal number of types). This second probability to be computed can represent the maximum value (i.e., the maximum quality evaluation score) under the browsing state features of the second sample user, jointly output by the quantitative prediction model and the quantitative evaluation model.

[0161] It is understood that the quantity prediction model in this application can be trained based on a reinforcement learning algorithm, and the reinforcement learning algorithm used in this application embodiment can be the QN (Q-Learning Network) algorithm. The training process of the quantity prediction model for reinforcement learning can be as follows:

[0162] It should be understood that the discount coefficient of the quantity prediction model can be obtained. Furthermore, the feedback parameter in the training sample data can be obtained (this feedback parameter can be used to evaluate the quality of the quantity of the first sample type in the training sample data). Then, based on the feedback parameter, the discount coefficient, and the second probability to be calculated, the target computational probability associated with the training sample data can be determined. It should be understood that the target computational probability can be understood as the actual quality score associated with the training sample data. At this point, the business server can determine the loss function value of the quantity prediction model based on the actual quality score and the first probability to be calculated (i.e., the predicted quality score).

[0163] The loss function of this quantitative model can be expressed as shown in formula (1):

[0164] loss = (γ × target) score +reward-eval score ) 2 Formula (1)

[0165] Here, γ can be used to characterize the discount factor of this quantity prediction model, typically taking a value between 0.9 and 0.99; target score It can be used to characterize the predicted quality score (second probability to be calculated) under the browsing state of the second sample user; evalscore It can be used to characterize the predicted quality score (i.e., the first probability to be calculated) under the browsing state of the first sample user; reward can be used to characterize the feedback parameter of the target user obtained after pushing public broadcast media data with the first sample type quantity under the browsing state of the first sample user; (γ×target) score +reward) can be used to characterize the actual quality score under the browsing state of the first sample user.

[0166] It should be understood that the loss function value can be obtained through the loss function shown in formula (1). The quantity prediction model can be trained based on the loss function value to obtain the target quantity prediction model. For example, it can be determined whether the loss function value meets the model convergence condition. If the loss function value does not meet the model convergence condition, the model parameters of the quantity prediction model can be optimized and adjusted based on the loss function value. If the loss function value meets the model convergence condition, it can be determined that the model is in a convergent state, and the quantity prediction model in a convergent state can be determined as the target quantity prediction model. The way to determine whether the loss function value meets the model convergence condition is as follows: it can be determined whether the number of model update iterations has reached the model update iteration threshold (if the model update iteration threshold is reached, it can be determined that the loss function value meets the model convergence condition), or whether the loss function value is within the error allowable range (if it is within the error allowable range, it can be determined that the loss function value meets the model convergence condition), and so on.

[0167] In this embodiment, the quantity prediction model is trained using the browsing status characteristics of the target user's sample users. This allows the trained model to accurately predict the first predicted quantity (the quantity of media data belonging to the first media type) for the target user under their current browsing status. Based on this first predicted quantity, a second predicted quantity (the quantity of media data belonging to the second media type) can be accurately determined. Public broadcast media data can then be pushed to the target user based on these two predicted quantities. In other words, this application can determine the ratio of media data belonging to the first media type to media data belonging to the second media type among the public broadcast media data pushed to the target user based on their current browsing status characteristics, thereby improving the accuracy of media data type recommendation.

[0168] To better understand the method for obtaining the target quantity prediction model from the training quantity prediction model, please also refer to... Figure 7 , Figure 7 This is a schematic diagram of a model training architecture provided in an embodiment of this application. For example... Figure 7As shown, the architecture can include a quantity prediction model and a quantity evaluation model. That is, the quantity prediction model can be trained using the quantity evaluation model. The quantity prediction model can be an action evaluation network (e.g., eval_net). The sample user browsing state features of the target user are input into this quantity prediction model, which outputs quality evaluation scores corresponding to N actions (in this application, actions represent configuration quantities, which can correspond to a first media type; if the first media type is video, the configuration quantity can be the number of video actions; if the first media type is text, the configuration quantity can be the number of text actions; if the first media type is music, the configuration quantity can be the number of music actions) under the sample user browsing state features. Thus, the quantity of the first sample type corresponding to the sample user browsing state features (which should be...) can be determined from the N quality evaluation scores. When the first media type is understood to be music, the configuration quantity can be the number of configured music tracks, and the first sample type quantity can be the number of sample music tracks. The quantity evaluation model can be a target evaluation network (e.g., target_net). Inputting the next sample user browsing state features of the target user into this quantity prediction model outputs a new action (i.e., the optimal type quantity). Inputting the next sample user browsing state features into the quantity prediction model outputs the quality evaluation score corresponding to the optimal type quantity. The loss function value can be determined by comparing the quality evaluation score of the first sample type quantity with the quality evaluation score corresponding to the optimal type quantity, and this loss function value is used to train the quantity prediction model. The quantity prediction model and the quantity evaluation model can have the same network structure; for example, the network structure can be a structure obtained by stacking multiple layers of a neural network, or it can be a structure consisting of only one fully connected layer. It should be understood that the training process for the quantity prediction model can be as follows:

[0169] First, the association information between the first sample user environment information and the first sample browsing behavior of the target user can be obtained. Based on this association information, the browsing state characteristics of the first sample user are determined. Then, the association information between the second sample user environment information and the second sample browsing behavior of the target user is obtained. Based on this association information, the browsing state characteristics of the second sample user are determined. For the specific implementation methods of determining the browsing state characteristics of the first and second sample users, please refer to the above. Figure 6 The description of step S201 in the corresponding embodiment will not be repeated here.

[0170] Furthermore, the browsing state features of the first sample user and the second sample user can be obtained, and training sample data for training the quantity prediction model can be constructed based on these features. The specific implementation method can be found above. Figure 2 The description of step S202 in the corresponding embodiment will not be repeated here.

[0171] Furthermore, the browsing state features of the first sample user in the training sample data can be input into the quantity prediction model. The quantity prediction model can output the predicted quality score Q1 (i.e., the first sample prediction probability) corresponding to the N configured quantities under the browsing state features of the first sample user. From the N predicted quality scores Q1, the quality evaluation score (i.e., the predicted quality score, eval_score) corresponding to the quantity of the first sample type in the training sample data can be selected. The quality evaluation score corresponding to the quantity of the first sample type is the first probability to be calculated mentioned above.

[0172] Furthermore, the browsing state features of the second sample user in the training sample data can be input into the quantity prediction model. The quantity prediction model can output the predicted quality score Q2 (i.e., the second sample prediction probability) corresponding to the N configuration quantities under the browsing state features of the second sample user. Among the N predicted quality scores Q2, the largest predicted quality score Q2 can be selected, and the configuration quantity corresponding to the largest predicted quality score Q2 can be determined as the optimal type quantity.

[0173] Furthermore, the browsing state features of the second sample user in the training sample data can be input into the quantity evaluation model. The quantity evaluation model can output the predicted quality score Q3 (i.e., the third sample prediction probability) corresponding to the N configuration quantities under the browsing state features of the second sample user. The predicted quality score corresponding to the optimal type quantity can be selected from the N predicted quality scores Q3, thereby obtaining the target quality evaluation score (such as target_score) corresponding to the optimal type quantity. The target quality evaluation score corresponding to the optimal type quantity is the second probability to be calculated mentioned above.

[0174] Furthermore, the feedback parameters corresponding to the number of sample videos can be determined through the browsing behavior association information of the second sample. Based on the feedback parameters, the first probability to be calculated, the second probability to be calculated, and the formula (1), a loss function value can be generated. The quantity prediction model can be trained and adjusted based on the loss function value to obtain the target quantity prediction model.

[0175] It should be understood that training the quantity prediction model using the browsing status characteristics of the target user's sample users enables the trained target quantity prediction model to accurately predict the first predicted quantity (the quantity of media data belonging to the first media type) corresponding to the target user's current browsing status characteristics. Therefore, based on the first predicted quantity, a second predicted quantity (the quantity of media data belonging to the second media type) can be accurately determined. Based on these first and second predicted quantities, public broadcast media data can be pushed to the target user. In other words, this application can determine the ratio of media data belonging to the first media type to media data belonging to the second media type among the public broadcast media data pushed to the target user based on the target user's current browsing status characteristics, thereby improving the accuracy of media data type recommendation.

[0176] Further, please see Figure 8 , Figure 8 This is a schematic diagram of the structure of a data processing apparatus provided in an embodiment of this application. The data processing apparatus can be a computer program (including program code) running on a computer device; for example, the data processing apparatus is an application software. The data processing apparatus can be used to execute... Figure 3 The method shown. (As illustrated) Figure 8 As shown, the data processing device 1 may include: an information acquisition module 11, a feature determination module 12, a quantity determination module 13, and a data push module 14.

[0177] The information acquisition module 11 is used to acquire user environment information corresponding to the target user and browsing behavior related information of the target user for the first public broadcast media data within the first time period.

[0178] The feature determination module 12 is used to determine the user browsing status features of the target user in relation to the first public broadcast media data based on user environment information and browsing behavior-related information.

[0179] The quantity determination module 13 is used to predict the first predicted quantity corresponding to the target user based on the user's browsing status characteristics; the first predicted quantity refers to the quantity of media data belonging to the first media type pushed to the terminal corresponding to the target user within the second time period; the first time period is earlier than the second time period;

[0180] The quantity determination module 13 is also used to determine the second predicted quantity based on the first predicted quantity; the second predicted quantity refers to the quantity of media data belonging to the second media type pushed to the terminal corresponding to the target user within the second time period;

[0181] The data push module 14 is used to push the second public broadcast media data to the terminal corresponding to the target user within a second time period based on the first predicted quantity and the second predicted quantity; the second public broadcast media data includes first target media data belonging to the first media type and second target media data belonging to the second media type; the quantity of the first target media data is the first predicted quantity; the quantity of the second target media data is the second predicted quantity.

[0182] The specific implementation methods of the information acquisition module 11, feature determination module 12, quantity determination module 13, and data push module 14 can be found in the above description. Figure 3 The descriptions of steps S101-S105 in the corresponding embodiments will not be repeated here.

[0183] In one embodiment, user environment information includes user attribute information and environment attribute information;

[0184] Please see Figure 8 The feature determination module 12 may include a hash value generation unit 121 and a feature determination unit 122.

[0185] The hash value generation unit 121 is used to generate a first hash value corresponding to the user attribute information based on the user attribute information and the hash function.

[0186] The hash value generation unit 121 is also used to generate a second hash value corresponding to the environmental attribute information based on the environmental attribute information and the hash function;

[0187] The hash value generation unit 121 is also used to generate a third hash value corresponding to the browsing behavior association information based on the browsing behavior association information and the hash function;

[0188] The feature determination unit 122 is used to determine the user browsing status features of the target user for the first public broadcast media data based on the first hash value, the second hash value and the third hash value.

[0189] The specific implementation methods of the hash value generation unit 121 and the feature determination unit 122 can be found in the above description. Figure 3 The description of step S102 in the corresponding embodiment will not be repeated here.

[0190] Please see Figure 8 The feature determination unit 122 may include: a table acquisition subunit 1221, a vector acquisition subunit 1222, and a vector fusion subunit 1223.

[0191] Table retrieval subunit 1221 is used to retrieve the vector mapping table; the vector mapping table includes the mapping relationship between configuration hash values ​​and configuration vectors; the configuration hash values ​​include a first hash value, a second hash value, and a third hash value;

[0192] The vector acquisition subunit 1222 is used to acquire a configuration vector that has a mapping relationship with the first hash value in the vector mapping table, and use it as the first user feature vector;

[0193] The vector acquisition subunit 1222 is also used to acquire, in the vector mapping table, a configuration vector that has a mapping relationship with the second hash value, as the second user feature vector;

[0194] The vector acquisition subunit 1222 is also used to acquire, in the vector mapping table, a configuration vector that has a mapping relationship with the third hash value, as the third user feature vector;

[0195] The vector fusion subunit 1223 is used to fuse the first user feature vector, the second user feature vector, and the third user feature vector to obtain the user browsing status features of the target user for the first public broadcast media data.

[0196] The specific implementation methods of the table acquisition subunit 1221, the vector acquisition subunit 1222, and the vector fusion subunit 1223 can be found in the above description. Figure 5 The descriptions in the corresponding embodiments will not be repeated here.

[0197] Please see Figure 8 The quantity determination module 13 may include: a probability prediction unit 131 and a first quantity determination unit 132.

[0198] The probability prediction unit 131 is used to input the user browsing status features into the target quantity prediction model, and output the prediction probabilities corresponding to N configuration quantities through the target quantity prediction model to obtain N prediction probabilities; each configuration quantity represents the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user in the second time period; N is an integer;

[0199] The first quantity determination unit 132 is used to obtain the maximum prediction probability among N prediction probabilities and determine the configuration quantity corresponding to the maximum prediction probability as the first prediction quantity.

[0200] The specific implementation methods of the probability prediction unit 131 and the first quantity determination unit 132 can be found in the above description. Figure 3 The description of step S103 in the corresponding embodiment will not be repeated here.

[0201] Please see Figure 8The quantity determination module 13 may include a quantity acquisition unit 133 and a second quantity determination unit 134.

[0202] The quantity acquisition unit 133 is used to acquire the total configured quantity; the total configured quantity refers to the total amount of media data required to be pushed to the terminal corresponding to the target user.

[0203] The second quantity determination unit 134 is used to subtract the total configuration quantity from the first predicted quantity to obtain the second predicted quantity.

[0204] The specific implementation methods of the quantity acquisition unit 133 and the second quantity determination unit 134 can be found in the above description. Figure 3 The description of step S104 in the corresponding embodiment will not be repeated here.

[0205] Please see Figure 8 The data push module 14 may include: a collection acquisition unit 141, a push data acquisition unit 142, and a data push unit 143.

[0206] The collection acquisition unit 141 is used to acquire a collection of media data to be pushed from the database. The collection of media data to be pushed includes media data of the first media type that is expected to be pushed to the terminal corresponding to the target user during the second time period, and media data of the second media type that is expected to be pushed to the terminal corresponding to the target user.

[0207] The push data acquisition unit 142 is used to acquire the first target media data from the media data set to be pushed according to the first predicted quantity;

[0208] The push data acquisition unit 142 is also used to acquire second target media data from the media data set to be pushed according to the second predicted quantity;

[0209] The data push unit 143 is used to determine the first target media data and the second target media data as the second public broadcast media data, and push the second public broadcast media data to the terminal corresponding to the target user within the second time period.

[0210] The specific implementation methods of the collection acquisition unit 141, the push data acquisition unit 142, and the data push unit 143 can be found in the above description. Figure 3 The description of step S105 in the corresponding embodiment will not be repeated here.

[0211] In this embodiment, user environment information corresponding to a target user within a first time period and browsing behavior association information of the target user regarding the first public broadcast media data can be obtained. Based on the user environment information and browsing behavior association information, the user browsing status characteristics of the target user (i.e., the target user's current preference status characteristics for different types of media data) can be determined. Based on the user browsing status characteristics, a first predicted quantity corresponding to the target user can be predicted. Based on the first predicted quantity, a second predicted quantity can be determined. Based on the first and second predicted quantities, the second public broadcast media data can be pushed to the target user within the second time period. The second public broadcast media data includes media data belonging to a first media type (i.e., the first target media data) and media data belonging to a second media type (i.e., the second target media data). The quantity of the first target media data is the first predicted quantity, and the quantity of the second target media data is the second predicted quantity. In other words, this application can predict, based on the user browsing status characteristics of the target user in the first time period, the number of media data belonging to the first media type and the number of media data belonging to the second media type when pushing second public broadcast media data to the target user in the second time period. Since the number of media data belonging to the first media type and the number of media data belonging to the second media type are predicted and determined based on the user browsing status characteristics (current preference status characteristics) of the target user, the number of media data belonging to the first media type and the number of media data belonging to the second media type are consistent with the target user's current preference status for the first and second media types, resulting in high accuracy. In summary, this application can improve the accuracy of media data type recommendation.

[0212] Further, please see Figure 9 , Figure 9 This is a schematic diagram of another data processing apparatus provided in an embodiment of this application. The data processing apparatus can be a computer program (including program code) running on a computer device; for example, the data processing apparatus is an application software. The data processing apparatus can be used to execute... Figure 6 The method shown. (As illustrated) Figure 9 As shown, the data processing device 2 may include: a feature acquisition module 21, a training sample determination module 22, and a model training module 23.

[0213] The feature acquisition module 21 is used to acquire the browsing status features of a first sample user and a second sample user associated with the target user. The browsing status features of the first sample user are determined based on the target user's first sample user environment information acquired within the first sample time period, and the target user's first browsing behavior related information regarding the first sample public broadcast media data. The browsing status features of the second sample user are determined based on the target user's second sample user environment information acquired within the second sample time period, and the target user's second browsing behavior related information regarding the second sample public broadcast media data. The first sample time period is earlier than the second sample time period.

[0214] The training sample determination module 22 is used to determine training sample data associated with the target user based on the browsing status characteristics of the first sample user and the browsing status characteristics of the second sample user.

[0215] The model training module 23 is used to train the quantity prediction model based on the training sample data to obtain the target quantity prediction model. The target quantity prediction model is used to predict the number of types corresponding to the browsing status characteristics of the target user. The number of types refers to the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user.

[0216] The specific implementation methods of the feature acquisition module 21, the training sample determination module 22, and the model training module 23 can be found in the above description. Figure 6 The descriptions of steps S201-S203 in the corresponding embodiments will not be repeated here.

[0217] Please see Figure 9 The feature acquisition module 21 may include: a sample information acquisition unit 211, a sample status determination unit 212, a sample quantity determination unit 213, and a sample feature determination unit 214.

[0218] The sample information acquisition unit 211 is used to acquire, within the first sample time period, the first sample user environment information corresponding to the target user, and the first sample browsing behavior association information of the target user for the first sample public broadcast media data.

[0219] The sample state determination unit 212 is used to determine the first sample user browsing state characteristics of the target user based on the first sample user environment information and the first sample browsing behavior association information.

[0220] The sample quantity determination unit 213 is used to select the target configuration quantity from N configuration quantities as the first sample type quantity corresponding to the first sample user browsing status characteristics; the first sample type quantity refers to the quantity of media data belonging to the first media type pushed to the terminal corresponding to the target user during the second sample time period.

[0221] The sample quantity determination unit 213 is also used to obtain the total configuration quantity; the total configuration quantity refers to the total amount of media data required to push to the terminal corresponding to the target user;

[0222] The sample feature determination unit 214 is used to determine the browsing status features of the second sample user based on the total number of configurations and the number of first sample types.

[0223] The specific implementation methods of the sample information acquisition unit 211, sample status determination unit 212, sample quantity determination unit 213, and sample feature determination unit 214 can be found in the above description. Figure 6 The description of step S201 in the corresponding embodiment will not be repeated here.

[0224] Please see Figure 9 The sample quantity determination unit 213 may include: a sample quantity determination subunit 2131, a sample data push subunit 2132, a sample data acquisition subunit 2133, and a sample feature determination subunit 2134.

[0225] The sample quantity determination subunit 2131 is used to subtract the total number of configurations from the number of the first sample type to obtain the number of the second sample type of media data belonging to the second media type pushed to the terminal corresponding to the target user during the second sample time period.

[0226] The sample data push subunit 2132 is used to push the second sample public broadcast media data to the terminal corresponding to the target user within the second sample time period according to the number of the first sample type and the number of the second sample type; the second sample public broadcast media data includes the first sample media data belonging to the first media type and the second sample media data belonging to the second media type; the number of the first sample media data is the number of the first sample type, and the number of the second sample media data is the number of the second sample type;

[0227] The sample data acquisition subunit 2133 is used to acquire the second sample user environment information corresponding to the target user and the second sample browsing behavior association information of the target user for the second sample public broadcast media data during the second sample time period.

[0228] The sample feature determination subunit 2134 is used to determine the second sample user status features of the target user based on the second sample user environment information and the second sample browsing behavior association information.

[0229] The specific implementation methods of the sample quantity determination subunit 2131, sample data push subunit 2132, sample data acquisition subunit 2133, and sample feature determination subunit 2134 can be found above. Figure 6The description of step S201 in the corresponding embodiment will not be repeated here.

[0230] In one embodiment, the second sample browsing behavior association information includes the sample click frequency of the target user on the second sample public broadcast media data;

[0231] Please see Figure 9 The training sample determination module 22 may include: parameter determination unit 221 and training sample determination unit 222.

[0232] The parameter determination unit 221 is used to determine the sample click frequency as a feedback parameter associated with the target user and the number of the first sample type;

[0233] The training sample determination unit 222 is used to determine the data group composed of the browsing status features of the first sample user, the number of first sample types, feedback parameters, and the browsing status features of the second sample user as training sample data associated with the target user.

[0234] The specific implementation methods of parameter determination unit 221 and training sample determination unit 222 can be found in the above description. Figure 6 The description of step S202 in the corresponding embodiment will not be repeated here.

[0235] Please see Figure 9 The model training module 23 may include: a feature input unit 231, a probability selection unit 232, an optimal quantity determination unit 233, and a model training unit 234.

[0236] The feature input unit 231 is used to input the first sample user browsing status feature in the training sample data into the quantity prediction model, and output the first sample prediction probability corresponding to the N configuration video quantities through the quantity prediction model and the first sample user browsing status feature, so as to obtain N first sample prediction probabilities.

[0237] The probability selection unit 232 is used to determine the first sample prediction probability corresponding to the number of first sample types in the training sample data from N first sample prediction probabilities, and use it as the first probability to be calculated.

[0238] The feature input unit 231 is also used to input the second sample user browsing status feature in the training sample data into the quantity prediction model, and output the second sample prediction probability corresponding to N configuration quantities through the quantity prediction model and the second sample user browsing status feature, so as to obtain N second sample prediction probabilities.

[0239] The optimal quantity determination unit 233 is used to obtain the maximum second sample prediction probability among N second sample prediction probabilities and determine the configuration quantity corresponding to the maximum second sample prediction probability as the optimal type quantity.

[0240] The feature input unit 231 is also used to input the browsing status features of the second sample user into the quantity evaluation model, and output the prediction probabilities of the third sample corresponding to N configuration quantities through the quantity evaluation model and the browsing status features of the second sample user, so as to obtain N prediction probabilities of the third sample.

[0241] The model training unit 234 is used to train the quantity prediction model based on the first probability to be calculated, the optimal number of types, the feedback parameters in the training sample data, and the prediction probabilities of N third samples, so as to obtain the target quantity prediction model.

[0242] The specific implementation methods of the feature input unit 231, probability selection unit 232, optimal quantity determination unit 233, and model training unit 234 can be found in the above description. Figure 6 The description of step S203 in the corresponding embodiment will not be repeated here.

[0243] Please see Figure 9 The model training unit 234 may include: a probability acquisition subunit 2341, a probability operation subunit 2342, a loss value generation subunit 2343, and a model training subunit 2344.

[0244] The probability acquisition subunit 2341 is used to acquire the third sample prediction probability corresponding to the optimal number of types from N third sample prediction probabilities, as the second probability to be calculated;

[0245] The probability operation subunit 2342 is used to obtain the discount coefficient of the quantity prediction model and determine the target operation probability associated with the training sample data based on the discount coefficient, feedback parameters and the second probability to be operated.

[0246] The loss value generation subunit 2343 is used to subtract the target operation probability from the first operation probability to obtain the operation result, and determine the loss function value of the quantity prediction model based on the operation result;

[0247] Model training subunit 2344 is used to train the quantity prediction model based on the loss function value to obtain the target quantity prediction model.

[0248] The specific implementation methods of the probability acquisition subunit 2341, probability calculation subunit 2342, loss value generation subunit 2343, and model training subunit 2344 can be found above. Figure 5 The description of step S203 in the corresponding embodiment will not be repeated here.

[0249] In this embodiment, the quantity prediction model is trained using the browsing status characteristics of the target user's sample users. This allows the trained model to accurately predict the first predicted quantity (the quantity of media data belonging to the first media type) for the target user under their current browsing status. Based on this first predicted quantity, a second predicted quantity (the quantity of media data belonging to the second media type) can be accurately determined. Public broadcast media data can then be pushed to the target user based on these two predicted quantities. In other words, this application can determine the ratio of media data belonging to the first media type to media data belonging to the second media type among the public broadcast media data pushed to the target user based on their current browsing status characteristics, thereby improving the accuracy of media data type recommendation.

[0250] Further, please see Figure 10 , Figure 10 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Figure 10 As shown above, Figure 8 Device 1 in the corresponding embodiment or Figure 9 The device 2 in the corresponding embodiment can be applied to the aforementioned computer device 1000. The computer device 1000 may include a processor 1001, a network interface 1004, and a memory 1005. Furthermore, the computer device 1000 also includes a user interface 1003 and at least one communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen and a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as at least one disk storage device. Optionally, the memory 1005 may also be at least one storage device located remotely from the aforementioned processor 1001. Figure 10 As shown, the memory 1005, which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a device control application.

[0251] exist Figure 10 In the computer device 1000 shown, the network interface 1004 provides network communication functionality; the user interface 1003 is mainly used to provide an input interface for the user; and the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve:

[0252] Within the first time period, acquire user environment information corresponding to the target user, as well as information related to the target user's browsing behavior on the first public broadcast media data;

[0253] Based on user environment information and browsing behavior-related information, determine the user browsing status characteristics of the target user for the first public broadcast media data;

[0254] The first predicted quantity is predicted based on the user's browsing status characteristics; the first predicted quantity refers to the number of media data belonging to the first media type pushed to the terminal corresponding to the target user during the second time period; the first time period is earlier than the second time period.

[0255] The second predicted quantity is determined based on the first predicted quantity; the second predicted quantity refers to the amount of media data belonging to the second media type pushed to the terminal corresponding to the target user within the second time period.

[0256] Based on the first predicted quantity and the second predicted quantity, the second public broadcast media data is pushed to the terminal corresponding to the target user within the second time period; the second public broadcast media data includes the first target media data belonging to the first media type and the second target media data belonging to the second media type; the quantity of the first target media data is the first predicted quantity; the quantity of the second target media data is the second predicted quantity.

[0257] Or achieve:

[0258] The browsing status features of a first sample user and a second sample user associated with the target user are obtained. The browsing status features of the first sample user are determined based on the target user's first sample user environment information obtained within the first sample time period, and the target user's first browsing behavior related information regarding the first sample public broadcast media data. The browsing status features of the second sample user are determined based on the target user's second sample user environment information obtained within the second sample time period, and the target user's second browsing behavior related information regarding the second sample public broadcast media data. The first sample time period is earlier than the second sample time period.

[0259] Based on the browsing status characteristics of the first sample user and the browsing status characteristics of the second sample user, training sample data associated with the target user is determined.

[0260] The quantity prediction model is trained based on the training sample data to obtain the target quantity prediction model. The target quantity prediction model is used to predict the number of types corresponding to the browsing status characteristics of the target user. The number of types refers to the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user.

[0261] It should be understood that the computer device 1000 described in the embodiments of this application can execute the foregoing text. Figure 3 or Figure 6 The description of the data processing method in the corresponding embodiment can also be performed as described above. Figure 8 The corresponding embodiment refers to the data processing device 1, or the foregoing Figure 9 The description of the data processing device 2 in the corresponding embodiments will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated here.

[0262] Furthermore, it should be noted that this application embodiment also provides a computer-readable storage medium, which stores a computer program executed by the aforementioned data processing computer device 1000. The computer program includes program instructions, and when the processor executes the program instructions, it can execute the aforementioned... Figure 3 or Figure 6 The description of the data processing method in the corresponding embodiments is already provided and will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the computer-readable storage medium embodiments related to this application, please refer to the description of the method embodiments of this application.

[0263] The aforementioned computer-readable storage medium can be an internal storage unit of the data processing apparatus or computer device provided in any of the foregoing embodiments, such as a hard disk or memory of the computer device. The computer-readable storage medium can also be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., provided on the computer device. Furthermore, the computer-readable storage medium can include both internal and external storage units of the computer device. The computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0264] One aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in one aspect of the embodiments of this application.

[0265] The terms "first," "second," etc., in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the term "comprising," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or modules, but may optionally include steps or modules not listed, or may optionally include other step units inherent to these processes, methods, apparatuses, products, or devices.

[0266] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0267] The methods and related apparatuses provided in this application are described with reference to the method flowcharts and / or structural diagrams provided in this application. Specifically, each block of the method flowchart and / or structural diagram, as well as combinations of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to create a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, generate instructions for implementing the process. Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 A process or multiple processes and / or structures illustrate the steps of the functions specified in one or more boxes.

[0268] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A data processing method, characterized in that, include: Within the first time period, acquire user environment information corresponding to the target user, as well as the browsing behavior association information of the target user regarding the first public broadcast media data; Based on the user environment information and the browsing behavior association information, the user browsing status characteristics of the target user in relation to the first public broadcast media data are determined; The user's browsing status features are used to predict the first predicted quantity corresponding to the target user; the first predicted quantity refers to the quantity of media data belonging to the first media type pushed to the terminal corresponding to the target user within the second time period. The first time period is earlier than the second time period; wherein, predicting the first predicted quantity corresponding to the target user based on the user browsing status features includes: inputting the user browsing status features into a target quantity prediction model, and outputting prediction probabilities corresponding to N configuration quantities through the target quantity prediction model to obtain N prediction probabilities; each configuration quantity represents the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user within the second time period; N is an integer; obtaining the maximum prediction probability among the N prediction probabilities, and determining the configuration quantity corresponding to the maximum prediction probability as the first predicted quantity; A second predicted quantity is determined based on the first predicted quantity; the second predicted quantity refers to the quantity of media data belonging to the second media type pushed to the terminal corresponding to the target user during the second time period; wherein, determining the second predicted quantity based on the first predicted quantity includes: obtaining a total configuration quantity; the total configuration quantity refers to the total quantity of media data required to be pushed to the terminal corresponding to the target user, and the total configuration quantity remains unchanged; subtracting the total configuration quantity from the first predicted quantity to obtain the second predicted quantity; Based on the first predicted quantity and the second predicted quantity, the second public broadcast media data is pushed to the terminal corresponding to the target user within the second time period; the second public broadcast media data includes first target media data belonging to the first media type and second target media data belonging to the second media type; the quantity of the first target media data is the first predicted quantity; the quantity of the second target media data is the second predicted quantity.

2. The method according to claim 1, characterized in that, The user environment information includes user attribute information and environment attribute information; The step of determining the user browsing status characteristics of the target user regarding the first public broadcast media data based on the user environment information and the browsing behavior association information includes: Based on the user attribute information and the hash function, a first hash value corresponding to the user attribute information is generated; Based on the environmental attribute information and the hash function, a second hash value corresponding to the environmental attribute information is generated; Based on the browsing behavior association information and the hash function, a third hash value corresponding to the browsing behavior association information is generated; Based on the first hash value, the second hash value, and the third hash value, the user browsing status characteristics of the target user regarding the first public broadcast media data are determined.

3. The method according to claim 2, characterized in that, The step of determining the user browsing status characteristics of the target user regarding the first public broadcast media data based on the first hash value, the second hash value, and the third hash value includes: Obtain a vector mapping table; the vector mapping table includes the mapping relationship between configuration hash values ​​and configuration vectors; the configuration hash values ​​include the first hash value, the second hash value, and the third hash value; In the vector mapping table, a configuration vector that has the mapping relationship with the first hash value is obtained as the first user feature vector; In the vector mapping table, a configuration vector that has the mapping relationship with the second hash value is obtained as the second user feature vector; In the vector mapping table, obtain the configuration vector that has the mapping relationship with the third hash value, and use it as the third user feature vector; The first user feature vector, the second user feature vector, and the third user feature vector are fused to obtain the user browsing status features of the target user in relation to the first public broadcast media data.

4. The method according to claim 1, characterized in that, The step of pushing the second public broadcast media data to the terminal corresponding to the target user within the second time period based on the first predicted quantity and the second predicted quantity includes: In the database, a set of media data to be pushed is obtained; the set of media data to be pushed includes media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user during the second time period, and media data belonging to the second media type that is expected to be pushed to the terminal corresponding to the target user. Based on the first predicted quantity, obtain the first target media data from the set of media data to be pushed; Based on the second predicted quantity, obtain the second target media data from the set of media data to be pushed; The first target media data and the second target media data are identified as the second public broadcast media data, and the second public broadcast media data is pushed to the terminal corresponding to the target user during the second time period.

5. A data processing method, characterized in that, include: Obtain the browsing status features of the first sample user and the browsing status features of the second sample user associated with the target user; The first sample user browsing status feature is determined based on the first sample user environment information of the target user obtained within the first sample time period, and the first browsing behavior association information of the target user for the first sample public broadcast media data; the second sample user browsing status feature is determined based on the second sample user environment information of the target user obtained within the second sample time period, and the second browsing behavior association information of the target user for the second sample public broadcast media data; the first sample time period is earlier than the second sample time period; Based on the browsing status characteristics of the first sample user and the browsing status characteristics of the second sample user, training sample data associated with the target user is determined; The quantity prediction model is trained based on the training sample data to obtain the target quantity prediction model; the target quantity prediction model is used to predict the number of types corresponding to the browsing status characteristics of the target user, and the number of types refers to the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user.

6. The method according to claim 5, characterized in that, The acquisition of the first sample user browsing status features and the second sample user browsing status features associated with the target user includes: Within the first sample time period, obtain the first sample user environment information corresponding to the target user, and the first sample browsing behavior association information of the target user for the first sample public broadcast media data; Based on the first sample user environment information and the first sample browsing behavior association information, the first sample user browsing status characteristics of the target user are determined; Among the N configuration quantities, the target configuration quantity is selected as the first sample type quantity corresponding to the browsing status characteristics of the first sample user; the first sample type quantity refers to the quantity of media data belonging to the first media type pushed to the terminal corresponding to the target user during the second sample time period; Get the total configuration quantity; the total configuration quantity refers to the total amount of media data required to be pushed to the terminal corresponding to the target user. Based on the total number of configurations and the number of the first sample types, the browsing status characteristics of the second sample user are determined.

7. The method according to claim 6, characterized in that, The step of determining the browsing status characteristics of the second sample user based on the total number of configurations and the number of the first sample types includes: Subtract the total number of configurations from the number of the first sample types to obtain the number of the second sample types of media data belonging to the second media type pushed to the terminal corresponding to the target user during the second sample time period; Based on the number of the first sample type and the number of the second sample type, the second sample public broadcast media data is pushed to the terminal corresponding to the target user within the second sample time period; the second sample public broadcast media data includes first sample media data belonging to the first media type and second sample media data belonging to the second media type; the number of the first sample media data is the number of the first sample type, and the number of the second sample media data is the number of the second sample type; During the second sample time period, the second sample user environment information corresponding to the target user and the second sample browsing behavior association information of the target user for the second sample public broadcast media data are obtained; Based on the second sample user environment information and the second sample browsing behavior association information, the second sample user state characteristics of the target user are determined.

8. The method according to claim 7, wherein the second sample browsing behavior association information includes the sample click frequency of the target user on the second sample public broadcast media data; The step of determining the training sample data associated with the target user based on the browsing state features of the first sample user and the browsing state features of the second sample user includes: The sample click frequency is determined as a feedback parameter associated with the target user and the number of the first sample type; The data set consisting of the browsing status features of the first sample user, the number of the first sample types, the feedback parameters, and the browsing status features of the second sample user is determined as the training sample data associated with the target user.

9. The method according to claim 8, characterized in that, The step of training the quantity prediction model based on the training sample data to obtain the target quantity prediction model includes: The first sample user browsing status feature in the training sample data is input into the quantity prediction model. The quantity prediction model and the first sample user browsing status feature are used to output the first sample prediction probability corresponding to the number of N configured videos, respectively, to obtain N first sample prediction probabilities. Among the N first sample prediction probabilities, the first sample prediction probability corresponding to the number of first sample types in the training sample data is determined as the first probability to be calculated. The second sample user browsing status feature in the training sample data is input into the quantity prediction model. The quantity prediction model and the second sample user browsing status feature are used to output the second sample prediction probability corresponding to the N configured quantities, respectively, to obtain N second sample prediction probabilities. Among the N second sample prediction probabilities, the highest second sample prediction probability is obtained, and the configuration number corresponding to the highest second sample prediction probability is determined as the optimal type number. The second sample user browsing status features are input into the quantity evaluation model. The quantity evaluation model and the second sample user browsing status features are used to output the third sample prediction probability corresponding to the N configuration quantities, respectively, to obtain N third sample prediction probabilities. The quantity prediction model is trained based on the first probability to be calculated, the number of optimal types, the feedback parameters in the training sample data, and the prediction probabilities of the N third samples to obtain the target quantity prediction model.

10. The method according to claim 9, characterized in that, The step of training the quantity prediction model based on the first probability to be calculated, the number of optimal types, the feedback parameters in the training sample data, and the prediction probabilities of the N third samples to obtain the target quantity prediction model includes: Among the N predicted probabilities of the third sample, the predicted probability of the third sample corresponding to the optimal number of types is obtained as the second probability to be calculated; Obtain the discount coefficient of the quantity prediction model, and determine the target operation probability associated with the training sample data based on the discount coefficient, the feedback parameter, and the second probability to be operated. The target probability is subtracted from the first probability to be calculated to obtain the calculation result, and the loss function value of the quantity prediction model is determined based on the calculation result. The quantity prediction model is trained based on the loss function value to obtain the target quantity prediction model.

11. A data processing apparatus, characterized in that, include: The information acquisition module is used to acquire user environment information corresponding to the target user and browsing behavior-related information of the target user for the first public broadcast media data within a first time period. The feature determination module is used to determine the user browsing status features of the target user in relation to the first public broadcast media data based on the user environment information and the browsing behavior association information. The quantity determination module is used to predict a first predicted quantity corresponding to the target user based on the user browsing status characteristics; the first predicted quantity refers to the quantity of media data belonging to a first media type pushed to the terminal corresponding to the target user within a second time period; The first time period is earlier than the second time period; wherein, predicting the first predicted quantity corresponding to the target user based on the user browsing status features includes: inputting the user browsing status features into a target quantity prediction model, and outputting prediction probabilities corresponding to N configuration quantities through the target quantity prediction model to obtain N prediction probabilities; each configuration quantity represents the number of media data belonging to the first media type that is expected to be pushed to the terminal corresponding to the target user within the second time period; N is an integer; obtaining the maximum prediction probability among the N prediction probabilities, and determining the configuration quantity corresponding to the maximum prediction probability as the first predicted quantity; The quantity determination module is further configured to determine a second predicted quantity based on the first predicted quantity; the second predicted quantity refers to the quantity of media data belonging to the second media type pushed to the terminal corresponding to the target user within the second time period; wherein, determining the second predicted quantity based on the first predicted quantity includes: obtaining a total configured quantity; the total configured quantity refers to the total quantity of media data required to be pushed to the terminal corresponding to the target user, and the total configured quantity remains unchanged; subtracting the total configured quantity from the first predicted quantity to obtain the second predicted quantity; The data push module is used to push the second public broadcast media data to the terminal corresponding to the target user within the second time period according to the first predicted quantity and the second predicted quantity; the second public broadcast media data includes first target media data belonging to the first media type and second target media data belonging to the second media type; the quantity of the first target media data is the first predicted quantity; the quantity of the second target media data is the second predicted quantity.

12. A computer device, characterized in that, include: Processor, memory, and network interface; The processor is connected to the memory and the network interface, wherein the network interface is used to provide network communication functions, the memory is used to store program code, and the processor is used to call the program code to execute the method according to any one of claims 1-10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and to execute the method according to any one of claims 1-10.

Images