Live broadcast data processing method and related device

Abstract

The invention provides a live broadcast data processing method and a related device. The embodiment of the invention can be applied to the field of artificial intelligence. The method comprises the following steps: determining a specified group according to a target live broadcast, wherein the specified group comprises a plurality of online objects associated with the target live broadcast; feature data of each online object in the specified group is acquired, the feature data is used for calculating a watching probability, and the watching probability indicates a probability of watching the target live broadcast; obtaining the watching probability of the online object through a neural network according to the feature data; determining the online object with the watching probability higher than a first preset value as a target object; and displaying the number of the target objects through terminal equipment corresponding to the target live broadcast. The method is used for helping anchors optimize live broadcast strategies.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a live broadcast data processing method and related devices. Background Art

[0002] With the widespread adoption of internet technology and smartphones, more and more users are using live streaming platforms for socializing, shopping, entertainment, and education, leading to rapid growth in the live streaming industry. In live streaming scenarios, the main participants are the host and the audience. The host creates a live streaming room using a corresponding host terminal, and the audience accesses the room through a corresponding audience terminal to watch.

[0003] In the live broadcast scenario, the host user and the audience user can interact through the live broadcast interface. The host can see the real-time interactions of the audience users in the live broadcast room, such as likes, messages, and gifts, as well as the total number of viewers in the current live broadcast room.

[0004] The current problem is that it is difficult for anchor users to optimize the current live broadcast strategy based solely on the audience situation in the live broadcast room. Summary of the Invention

[0005] The embodiments of the present application provide a live broadcast data processing method and related devices, which help the anchor optimize the live broadcast strategy by showing the number of viewers who may enter the live broadcast room to the anchor.

[0006] The first aspect of the present application provides a live broadcast data processing method, comprising:

[0007] Determine a designated group according to the target live broadcast, the designated group including multiple online objects associated with the target live broadcast;

[0008] Obtaining feature data of each online subject in a specified group, the feature data being used to calculate a viewing probability, where the viewing probability indicates the probability of viewing a target live broadcast;

[0009] Obtain the viewing probability of an online object through a neural network based on feature data;

[0010] Determine an online object whose viewing probability is higher than a first preset value as a target object;

[0011] The number of target objects is displayed through the terminal device corresponding to the target live broadcast.

[0012] A second aspect of the present application provides a neural network training method, comprising:

[0013] Determine the training object associated with the target live broadcast;

[0014] Acquire feature data of a training object, the feature data being used to calculate a viewing probability, where the viewing probability indicates a probability of viewing a target live broadcast;

[0015] Obtain the predicted probability of the training object through the neural network based on the feature data;

[0016] The neural network is trained based on a loss function that indicates the similarity between the predicted probability and the expected probability, which indicates how often the training subjects actually watched the target live broadcast.

[0017] In one possible implementation method, the feature data includes the ranking of the target live broadcast in the waiting-to-be-played queue of the training object, the exposure of the target live broadcast, and at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast and / or similar live broadcasts. Similar live broadcasts indicate live broadcasts of the same type as the target live broadcast.

[0018] A third aspect of the present application provides a live broadcast data processing device, comprising:

[0019] A first determining module is configured to determine a designated group according to a target live broadcast, where the designated group includes a plurality of online objects associated with the target live broadcast;

[0020] The first acquisition module is further configured to acquire feature data of each online subject in a specified group, the feature data being used to calculate a viewing probability, where the viewing probability indicates a probability of viewing a target live broadcast;

[0021] A first prediction module, configured to obtain a viewing probability of an online object through a neural network based on feature data;

[0022] A processing module, configured to determine an online object with a viewing probability higher than a first preset value as a target object;

[0023] The output module is used to display the number of target objects through the terminal device corresponding to the target live broadcast.

[0024] In a possible implementation method, the multiple online objects include an active object of the target live broadcast, and the characteristic data of the active object includes at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast.

[0025] In a possible implementation method, the plurality of online objects include a first delivered object, a to-be-played queue of the first delivered object includes at least one to-be-played live broadcast, and the target live broadcast belongs to the to-be-played live broadcast.

[0026] In an optional case of a possible implementation method, the characteristic data of the first delivered object includes the order of the target live broadcast in the queue to be played.

[0027] In an alternative case of a possible implementation method,

[0028] The output module is also used to determine the order of the target live broadcast in the queue to be played based on the first delivered object; determine the first delivered object corresponding to the target live broadcast whose order is less than the second preset value as the object to be played; and display the number of objects to be played through the terminal device.

[0029] In a possible implementation method, the plurality of online objects include a second delivered object, the information flow interface of the second delivered object includes at least one exposed live broadcast, and the target live broadcast belongs to the exposed live broadcast.

[0030] In an alternative case of a possible implementation method,

[0031] The characteristic data of the second delivery object includes at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the same type of live broadcast. The same type of live broadcast indicates the same type of live broadcast as the target live broadcast.

[0032] In an alternative case of a possible implementation method,

[0033] The output module is also used to display the number of the second delivered objects through the terminal device.

[0034] In one possible implementation,

[0035] The output module is also used to display the target object through the terminal device when the target object watches the target live broadcast.

[0036] A fourth aspect of the present application provides a neural network training device, comprising:

[0037] A second determination module is used to determine a training object associated with the target live broadcast;

[0038] A second acquisition module is used to acquire feature data of the training object, the feature data is used to calculate the viewing probability, and the viewing probability indicates the probability of watching the target live broadcast;

[0039] The second prediction module is used to obtain the predicted probability of the training object through a neural network based on the feature data;

[0040] A training module is used to train the neural network according to a loss function, where the loss function indicates the similarity between the predicted probability and the expected probability, and the expected probability indicates the actual situation of the training subject watching the target live broadcast.

[0041] In one possible implementation method, the feature data includes the ranking of the target live broadcast in the waiting-to-be-played queue of the training object, the exposure of the target live broadcast, and at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast and / or similar live broadcasts. Similar live broadcasts indicate live broadcasts of the same type as the target live broadcast.

[0042] A fifth aspect of the present application provides a computer device, comprising: a memory and a processor;

[0043] The memory stores instructions, and when the instructions are executed on the processor, the above methods are executed.

[0044] In a sixth aspect, the present application provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the methods in the above aspects.

[0045] In a seventh aspect, the present application provides a computer program product or computer program, the 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 the processor executes the computer instructions, causing the computer device to perform the methods provided in the above aspects.

[0046] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0047] The present application provides a live broadcast data processing method and related devices. First, a designated group is determined based on the target live broadcast. The designated group indicates the group of online objects associated with the target live broadcast, wherein association with the target live broadcast means that the online object has the possibility of entering the target live broadcast room, for example, the online object frequently visits the target live broadcast room, the online object is served with advertisements for the target live broadcast room, the target live broadcast room appears in the information interface of the online object, etc. After determining the designated group, characteristic data of each online object is obtained, and the characteristic data user calculates the probability of the corresponding online object watching the target live broadcast; based on the characteristic data, the viewing probability is calculated through a neural network, and the online object with a probability greater than a preset value is determined as the target object; the number of target objects, that is, the estimated number of viewers of the live broadcast room, is displayed to the terminal device corresponding to the target live broadcast. Through the above method, the anchor can determine the number of viewers who may currently enter the live broadcast room in real time during the live broadcast, and thus make corresponding live broadcast strategy adjustments. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A schematic diagram of the live broadcast interface for a host user in a live broadcast scenario;

[0049] Figure 2This is an application environment diagram of the live broadcast data processing method in an embodiment of the present application;

[0050] Figure 3 A flow chart of the live broadcast data processing method provided in an embodiment of the present application;

[0051] Figure 4a and Figure 4b A schematic diagram of the live broadcast interface of the anchor object provided in an embodiment of the present application;

[0052] Figure 5 A flow chart of the live broadcast data processing method provided in an embodiment of the present application;

[0053] Figure 6 Schematic diagram of a live streaming scenario;

[0054] Figure 7 Schematic diagram of several information flow interfaces including live broadcast entrances;

[0055] Figure 8 A schematic diagram of the interface for a target object to enter a live broadcast room provided in an embodiment of the present application;

[0056] Figure 9 A schematic diagram of the live broadcast interface of the anchor object provided in an embodiment of the present application;

[0057] Figure 10 This is a flowchart corresponding to the live data processing in an embodiment of the present application;

[0058] Figure 11a A flowchart of a neural network training method provided in an embodiment of the present application;

[0059] Figure 11b A training flow chart of a neural network training method;

[0060] Figure 12 A flow chart of the training of the logistic regression model provided in the embodiment of the present application;

[0061] Figure 13 This is a schematic diagram of an embodiment of a live broadcast data processing device in an embodiment of the present application;

[0062] Figure 14 This is a schematic diagram of an embodiment of a neural network training device in an embodiment of the present application;

[0063] Figure 15 This is a schematic diagram of a server structure provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the numbers used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can, for example, be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "corresponding to" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0065] Artificial Intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to 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 seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. AI also involves studying the design principles and implementation methods of various intelligent machines, enabling them to possess the capabilities of perception, reasoning, and decision-making.

[0066] Artificial intelligence (AI) technology is a comprehensive discipline encompassing a wide range of fields, encompassing both hardware and software technologies. Foundational AI technologies generally include sensors, specialized AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0067] Machine learning (ML) is a multidisciplinary field that encompasses probability theory, statistics, approximation theory, convex analysis, and algorithmic complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is at the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications span all areas of AI. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0068] The solution provided in the embodiments of the present application involves machine learning technology. By constructing training data to train a prediction model, the prediction model can predict the probability of audience users entering a target live broadcast room. Based on the prediction results, the number of viewers who may enter the target live broadcast room is counted, so that the anchor user can adjust the live broadcast strategy based on the number of viewers. The specific details are explained in the embodiments below.

[0069] Next, we introduce the technical background of the live broadcast data processing method and related devices provided in the embodiments of the present application:

[0070] With the widespread adoption of internet technology and smartphones, more and more users are using live streaming platforms for socializing, shopping, entertainment, and education, leading to rapid growth in the live streaming industry. In live streaming scenarios, the main participants are the host and the audience. The host creates a live streaming room using a corresponding host terminal, and the audience accesses the room through a corresponding audience terminal to watch.

[0071] With the widespread adoption of internet technology and smartphones, the live streaming industry has experienced rapid growth. Live streaming has become a popular way to socialize, shop, entertain, and educate, attracting a large number of users. In live streaming scenarios, the main participants are the host and the audience.

[0072] Livestreamers are those who create livestreaming rooms on livestreaming platforms and broadcast live. They typically use specialized livestreaming software or platform-provided tools to broadcast livestreams via devices such as smartphones and computers. Livestreamers can choose different types of livestreams based on their interests and specialties, such as gaming, lifestyle, and beauty.

[0073] Audience users are those who access the live broadcast room through their respective terminals and watch the live broadcast. They can watch the live broadcast on various devices, such as smartphones, tablets, and TVs. They can communicate and interact with the host through the interactive features provided by the live broadcast platform, such as sending comments, liking, leaving messages, and sending gifts.

[0074] See also Figure 1 , Figure 1 A schematic diagram of the live broadcast interface of an anchor user in a live broadcast scenario.

[0075] The live broadcast interface of the anchor user mainly includes the following parts: anchor information 101, live broadcast status 102, interactive window 103 and live broadcast options 104.

[0076] The host information 101 includes the host's profile picture and the number of likes for the live broadcast. The higher the number of likes, the stronger the host's interactive ability and the better the live broadcast effect.

[0077] The live broadcast status 102 includes the number of viewers in the live broadcast room and the network status. The more viewers there are, the more popular and influential the anchor is. This indicator is a key factor in advertising opportunities.

[0078] Interaction window 103 displays the interactions between audience users and the host user, including conversations, new users, and gift-giving interactions. As shown, conversations include: "Zhang San: Finally here!", "Li Si: Here's the link!", and "Zhou Ba: The big boss is here, hahaha!"; new users include: "Super fan Wang Wu enters the live broadcast room!"; and gift-giving interactions include: "Sun Qi sends the host five flowers." It is understandable that Zhang San, Li Si, Wang Wu, Zhao Liu, Sun Qi, and Zhou Ba are all audience users.

[0079] The live broadcast options 104 include live broadcast options that the anchor can click, for example: click the "Interaction" option to perform operations such as sending red envelopes, playing games, and drawing prizes to interact with the audience; click the "Decoration" option to add filters, stickers, special effects and other decorative effects to the live broadcast; click the "End Broadcast" option to end the live broadcast.

[0080] In the live streaming scenario described above, the host can see real-time audience interactions, such as likes, comments, and spending, as well as the total number of viewers in the current live streaming room, through the live streaming interface. However, the current problem is that it is difficult for the host to optimize the current live streaming strategy based solely on the audience data in the live streaming room.

[0081] Based on the above problems, an embodiment of the present application provides a live broadcast data processing method and related devices, which estimates the number of new viewers who may enter the live broadcast room and displays it to the anchor, so that the anchor can adjust the current live broadcast strategy according to the estimated number of new viewers.

[0082] For easier understanding, see Figure 2 , Figure 2 This is an application environment diagram of the live broadcast data processing method in the embodiment of the present application, such as Figure 2As shown, the live broadcast data processing method in the embodiment of the present application is applied to a live broadcast data processing system. The live broadcast data processing system includes: a server and a terminal device; wherein the server can be an independent physical server, or 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 communications, middleware services, domain name services, security services, content distribution networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms. The terminal device can be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited to this. The terminal device and the server can be directly or indirectly connected by wired or wireless communication, and the embodiment of the present application is not limited here.

[0083] This method is applied to a live broadcast scenario. It is used to estimate the number of new viewers who may enter the live broadcast room and display it to the host. The following describes this method from the perspective of the server:

[0084] The server first determines a designated group based on the target live broadcast, where the designated group indicates a group of online objects associated with the target live broadcast, wherein association with the target live broadcast means that the online object has the possibility of entering the target live broadcast room, for example, the online object frequently visits the target live broadcast room, the online object is served advertisements for the target live broadcast room, the target live broadcast room appears in the information interface of the online object, etc.; after determining the designated group, the server obtains feature data of each online object, and uses the feature data to calculate the probability of the corresponding online object watching the target live broadcast; based on the feature data, the viewing probability is calculated through a neural network, and online objects with a probability greater than a preset value are determined as target objects; and the number of target objects, that is, the estimated number of viewers of the live broadcast room, is displayed to the terminal device corresponding to the target live broadcast.

[0085] It is understandable that the above method is executed by a computer device, which may be Figure 2 In this embodiment, the live data processing method in this application will be introduced from the perspective of the computer device as the server. Figure 3 The live broadcast data processing method provided in the embodiment of the present application includes:

[0086] 301 : Determine a designated group according to a target live broadcast, where the designated group includes multiple online objects associated with the target live broadcast.

[0087] It is understood that the target live broadcast refers to the target live broadcast room for which the audience number estimation is required, and the audience number refers to the total number of viewers who may currently enter the target live broadcast room to watch. To estimate the total number of viewers, it is first necessary to determine a designated group, which includes multiple online objects associated with the target live broadcast.

[0088] Online objects refer to objects or entities that are using the software corresponding to the live broadcast platform, and are currently connected to the Internet network and are active registered objects. Online objects have established a real-time connection with the software and can interact with the software or with other objects through the software. For example, for social software A that can provide live broadcast functions, the live broadcast function is installed on live broadcast platform A. Through social software A, users can interact with other users in a one-on-one chat through dialog boxes in the form of text, pictures, videos, etc.; they can also publish articles, links, pictures and other shared content through the social platform provided by social software A; they can also watch live broadcasts on live broadcast platform B, etc. So as long as the user is using the social software, whether chatting, publishing and sharing content or watching live broadcasts, the user is an online object.

[0089] Based on the above definition of online objects, what the embodiment of the present application needs to obtain is the online object associated with the target live broadcast, that is, the online object has the possibility of entering the target live broadcast room, for example, the online object frequently visits the target live broadcast room, the online object is served advertisements of the target live broadcast room, the target live broadcast room appears in the information interface of the online object, etc.

[0090] 302 , obtaining feature data of each online object in a specified group, where the feature data is used to calculate a viewing probability, where the viewing probability indicates a probability of viewing a target live broadcast.

[0091] After multiple online objects are determined, feature data of each online object is obtained. The feature data indicates behavioral features of the online object in multiple dimensions and is used to calculate the probability of watching the target live broadcast.

[0092] For each online subject, feature data can be obtained through the online subject's live broadcast viewing history, including, but not limited to: the time interval between the subject's last live broadcast viewing, the subject's average live broadcast viewing time, the subject's current network status, the frequency of the subject's live broadcast viewing within 7 days / 21 days, whether the subject has sent gifts while watching live broadcasts within 7 days / 21 days, whether the subject has liked or commented on live broadcasts within 7 days / 21 days, etc. It is understandable that the above-mentioned live broadcasts can include the target live broadcast and other live broadcasts of the same type as the target live broadcast. Feature data can also be analyzed based on the online subject's browsing behavior, search behavior, etc. in the software, and the specific indicators of the feature data can be determined based on the experience of those skilled in the art.

[0093] 303 , obtaining the viewing probability of the online object through a neural network according to the feature data.

[0094] It is understood that neural networks can predict outcomes by learning from a large number of data samples. In the scenario corresponding to the embodiments of this application, the neural network can learn how to predict the probability of users watching live broadcasts based on their characteristic data. By training the neural network, we can enable it to understand the correlation between different characteristics and viewing behavior, and based on these correlations, predict the user's viewing probability.

[0095] Specifically, we can pre-acquire the feature data of sample subjects and their viewing results of the target live broadcast (whether they watched or not, represented by 1 or 0). This feature data of the sample subjects is then input into the neural network as training data. The neural network is then adjusted and optimized using the actual viewing results as expected values to obtain a trained neural network. By inputting the feature data of online subjects into the trained neural network, the viewing probability of the online user is obtained.

[0096] 304 : Determine an online object whose viewing probability is higher than a first preset value as a target object.

[0097] It can be understood that the viewing probability can be a value between 0 and 1. The higher the value, the higher the probability that the user will watch the target live broadcast. When this value is higher than a first preset value, the corresponding online object is determined as the target object. The target object refers to the audience object whose predicted result is to watch the target live broadcast; correspondingly, if this value is not higher than the first preset value, the corresponding online object is predicted to not watch the target live broadcast.

[0098] The viewing probability can also be represented by only 0 and 1, where 0 indicates not watching the target live broadcast, and 1 indicates watching the target live broadcast, that is, the value range of the first preset value is [0, 1).

[0099] 305 , displaying the number of target objects through the terminal device corresponding to the target live broadcast.

[0100] It can be understood that by predicting the viewing probability of each online subject, the number of target subjects in a specified group can be obtained, which is the estimated number of new viewers for the target live broadcast. The estimated number of new viewers is displayed on the terminal device corresponding to the target live broadcast, so that the host can adjust the live broadcast strategy based on the estimated result.

[0101] For easier understanding, see Figure 4a and Figure 4b , Figure 4a and Figure 4b Schematic diagram of the live broadcast interface of the anchor object provided in an embodiment of the present application.

[0102] Based on the method provided in the embodiment of the present application, the estimated number of new viewers (the number of target viewers) can be viewed at any time in the live broadcast interface of the host object.

[0103] like Figure 4a As shown in the above Figure 1 On the basis of the corresponding live interface diagram, a new to-be-played option 401 is added to the live option 104. In response to the host object clicking the to-be-played option 401, the Figure 4b The interface is shown. Figure 4b As shown, an information box appears on the original live broadcast interface, and the information box is used to display the estimated number of new people (the number of target objects) obtained based on the live broadcast data processing method in the above embodiment.

[0104] The following are several possible scenarios for adjusting the live broadcast strategy based on the live broadcast data processing method provided in this application:

[0105] Scenario 1: Determine the next broadcast content based on the estimated number of new viewers: The broadcaster can choose appropriate topics or content based on the estimated number of new viewers. For example, if the expected number of new viewers is high, the broadcaster can promote a popular product to appeal to the interests of the majority of viewers and increase live broadcast revenue. If the expected number of new viewers is low, the broadcaster can increase interaction with the audience and increase the popularity of the live broadcast room to gain more exposure opportunities.

[0106] Case 2: Determine whether to run promotions based on the estimated number of new viewers: If the number of new viewers expected is small, consider running promotions to attract more viewers and increase exposure for the livestream. If the number of new viewers is already large, the streamer does not need to run promotions, but instead focuses on ensuring the quality of the current livestream and audience engagement.

[0107] Scenario 3: Determine broadcast duration based on estimated audience: Based on the estimated number of viewers, the broadcaster can determine the appropriate broadcast duration. If the estimated audience is high, the broadcaster can extend the current broadcast duration and increase interaction to meet audience needs, thereby increasing popularity and revenue. If the estimated audience is low, the broadcaster can end the broadcast early, eliminating the need to waste time waiting and saving energy for subsequent broadcasts.

[0108] The live broadcast data processing method provided in the embodiments of the present application first determines the group of online subjects associated with the target live broadcast, then obtains the characteristic data of each online subject in the group. This characteristic data is then predicted using a pre-trained neural network to obtain the probability of each online subject viewing the target live broadcast. Based on this probability, the target subjects who will watch the target live broadcast are determined, and finally the number of target subjects is displayed on the terminal device corresponding to the target live broadcast. This method allows the host to determine the number of viewers who may enter the live broadcast room in real time during the live broadcast, thereby adjusting the live broadcast strategy accordingly.

[0109] In this application Figure 3 In an optional embodiment of the live data processing method provided by the corresponding embodiment, please refer to Figure 5 , Figure 5 The method flow chart of the live broadcast data processing method provided in the embodiment of the present application includes:

[0110] 501. Determine a designated group according to a target live broadcast, where the designated group includes multiple online objects associated with the target live broadcast.

[0111] It is understandable that step 501 in the embodiment of the present application is the same as the above Figure 3 The step 301 in the corresponding embodiment is similar.

[0112] In the embodiments of the present application, the following possible implementation methods are provided for introducing the types of online objects included in a specified group, which are described below:

[0113] In a possible implementation method, multiple online objects of a designated group include active objects of a target live broadcast.

[0114] Active users of the target live broadcast refer to users who are active in the live broadcast room corresponding to the target live broadcast and therefore have the possibility of watching the target live broadcast. Active users can be determined based on the viewing time, interactive behavior, and consumption behavior of the target live broadcast:

[0115] Watching time: Active subjects may spend a long time watching the target live broadcast. Therefore, the length of time an online subject watches the target live broadcast can be used to measure whether they are active. That is, online subjects who watch the target live broadcast for more than a preset time within a certain time range, for example, online subjects who watch the target live broadcast for more than 5 hours within 7 days.

[0116] Interactive behavior: Active subjects may interact with other subjects or the host in the target live broadcast, including posting comments, liking, sending gifts, and other behaviors. Therefore, you can define online subjects as those who interact with the target live broadcast more than a preset number of times within a certain time frame as active subjects, for example, online subjects who comment / like / send gifts more than 50 times in 7 days.

[0117] Consumption behavior: Active subjects may make purchases during the target live broadcast, such as purchasing products in the live broadcast room. Therefore, online subjects who purchase products during the target live broadcast more than a preset number of times within a certain time frame can be defined as active subjects, for example, online subjects who purchase products more than once within 7 days.

[0118] In another possible implementation method, the plurality of online objects of the designated group include a first delivered object, a to-be-played queue of the first delivered object includes at least one to-be-played live broadcast, and the target live broadcast belongs to the to-be-played live broadcast.

[0119] The waiting-to-play queue refers to the scene where live broadcasts or videos to be played are stored in a queue in a certain order while online objects are watching live broadcasts or videos, and are waiting to be played. The live broadcasts in the waiting-to-play queue are waiting-to-play live broadcasts. Figure 6 As shown, Figure 6 This is a schematic diagram of a live streaming scenario. An online user is watching Live Broadcast N of Host 1 on their mobile phone. They can also swipe up on the screen to access Live Broadcast N+1 of Host 2. In response to the online user's swipe up action, Live Broadcasts N+1 and N+2 presented are both pending live broadcasts. This sequentially arranged list of live broadcasts is the pending broadcast queue.

[0120] It is understandable that if the target live broadcast is in the to-be-played list of the online object, then the target live broadcast is very likely to be watched by the online object, so the online objects associated with the target live broadcast include the first delivered object.

[0121] In another possible implementation method, the plurality of online objects of the designated group include a second delivered object, the information flow interface of the second delivered object includes at least one exposed live broadcast, and the target live broadcast belongs to the exposed live broadcast.

[0122] An information flow interface refers to an interface that presents various types of information to users in a stream manner. Various types of information are organized into a series of cards and modules. Each card and module includes information such as articles, pictures, and videos. Users can browse these cards or modules by sliding or scrolling. Figure 7 As shown, Figure 7 This is a schematic diagram of several information flow interfaces including live broadcast entrances, including the live broadcast square interface, chat interface and search interface.

[0123] In the Live Square interface, live stream cards for each live stream are displayed directly on the interface. In the chat interface, other users can share live stream cards with online users through the chat box. In the search interface, live stream cards that meet the search criteria are displayed to online users. Live streams displayed in these interfaces are exposed live streams, and online users can click on the live stream card of their interest to watch the corresponding live stream.

[0124] It is understandable that in the scenario corresponding to the information flow interface, all types of live broadcasts generally go through the process of exposure, clicks, and viewing. Therefore, the exposed live broadcasts in the information flow interface have the possibility of being watched. Therefore, the online objects including the target live broadcast in the information flow interface include the second delivered objects.

[0125] In practical applications, the types of the above-mentioned online objects can be arbitrarily combined, that is, a designated group can include one or more types of the above-mentioned online objects.

[0126] 502, obtaining feature data of each online subject in a specified group, the feature data being used to calculate a viewing probability, where the viewing probability indicates a probability of viewing a target live broadcast;

[0127] It is understandable that step 502 in the embodiment of the present application is similar to the above Figure 3 Step 302 in the corresponding embodiment is similar.

[0128] exist Figure 3 In the case of the description of the characteristic data indicators in step 302 of the embodiment, specific characteristic data indicators may be added for the types of the online objects in step 501, which are described below:

[0129] When the online objects include active objects, the characteristic data of the active objects include at least one of viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast.

[0130] It is understandable that the active object has watched the target live broadcast before, so the probability of the active object watching the target live broadcast can be predicted based on the active object's viewing behavior of the target live broadcast.

[0131] When the online objects include the first delivered object, the characteristic data of the first delivered object includes the order of the target live broadcast in the queue to be played.

[0132] It is understandable that since the first delivery target includes the target live broadcast in the to-be-played list, the higher the ranking of the target live broadcast in the to-be-played list, the greater the possibility of it being played. Therefore, the probability of the first delivery target watching the target live broadcast can be predicted based on the ranking of the target live broadcast in the to-be-played list.

[0133] When a second delivery object is included, the characteristic data of the second delivery object includes at least one of viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to similar live broadcasts, and similar live broadcasts indicate live broadcasts of the same type as the target live broadcast.

[0134] It is understandable that the second recipient needs to click on the corresponding live broadcast card to watch the target live broadcast. Therefore, the higher the second recipient's interest in the target live broadcast, the greater the probability of playing the target live broadcast. To determine the second recipient's interest in the target live broadcast, the second recipient's viewing behavior for similar live broadcasts can be used to determine.

[0135] It should be noted that the characteristic data of the above-mentioned types of online objects include but are not limited to the indicators described above. For example, for an active object, its characteristic data may also include the ranking of the target live broadcast in the to-be-played list and the viewing behavior of similar live broadcasts; for the first delivered object, its characteristic data may also include the viewing behavior of the target live broadcast and the viewing behavior of similar live broadcasts; for the second delivered object, its characteristic data may also include the viewing behavior of the target live broadcast.

[0136] 503, obtaining a viewing probability of the online object through a neural network according to the feature data;

[0137] 504, determining an online object whose viewing probability is higher than a first preset value as a target object;

[0138] 505 , displaying the number of target objects through the terminal device corresponding to the target live broadcast.

[0139] It is understandable that steps 503 to 505 in the embodiment of the present application are similar to those in the above embodiment. Figure 3 Steps 303 to 305 in the corresponding embodiment are similar and will not be described again here.

[0140] Furthermore, it also includes:

[0141] 506. When the target object watches the target live broadcast, the target object is displayed through the terminal device.

[0142] In this embodiment, after the number of target objects is displayed to the terminal device corresponding to the target live broadcast, when the target objects actually watch the target live broadcast, the viewing status of the target objects can be displayed to the anchor to improve the anchor's trust in the estimated results of the live broadcast platform.

[0143] like Figure 8 As shown, Figure 8Schematic diagram of the interface for the target object to enter the live broadcast room provided in the embodiment of the present application. The audience user named "Wu Jiu" is the target object. When he enters the live broadcast room, the text of the newly added prompt can be changed by changing the color and font of the username to indicate to the anchor that the user is a predicted new user. In addition, the "To Play" button can be made to flash to prove that a predicted new user has entered the live broadcast room. It is understandable that those skilled in the art can show the target object watching the target live broadcast to the anchor in various ways, and this application does not limit the form of its display.

[0144] In a possible implementation method, after determining the first delivery target, the method further includes:

[0145] 507, based on the first delivered object, determining the order of the target live broadcast in the queue to be played;

[0146] 508, determining the first delivered object corresponding to the target live broadcast whose ranking is less than the second preset value as the object to be played;

[0147] 509. Display the number of objects to be played via the terminal device.

[0148] It can be understood that since the target live broadcast is included in the queue of the first delivered object, the ranking of the target live broadcast in the queue can be obtained; the smaller the ranking, the higher the possibility of the target live broadcast being played, so the objects to be played can be further screened out from the first delivered object; displaying the number of objects to be played by the terminal device can help the anchor subjectively estimate the future number of viewers, and further assist the anchor in adjusting the live broadcast strategy.

[0149] In a possible implementation method, after determining the second delivery target, the method further includes:

[0150] 510. Display the number of the second delivered objects through the terminal device.

[0151] It's understandable that the target live stream is exposed in the information flow interface of the second delivery target. Therefore, displaying the number of second delivery targets on the terminal device can help the live streamer confirm the current traffic status. For example, when the live streamer promotes the media platform to increase the popularity of the live broadcast room, the number of second delivery targets can reflect the promotion effect to a certain extent: the larger the number, the greater the platform's investment and the better the effect.

[0152] For easier understanding, see Figure 9 , Figure 9 Schematic diagram of the live broadcast interface of the anchor object provided in an embodiment of the present application.

[0153] It is understandable that the live interface can be clicked Figure 4a In the interface shown, after the "To Play" option 401, an information box appears on the original live broadcast interface. This information box includes not only the estimated number of new users, but also the top three users in the playlist and the number of users who have been exposed. The top three users in the playlist refer to the number of objects to be played in step 508 or 509, corresponding to the second preset value of 4; the number of users who have been exposed refers to the number of the second objects delivered in step 510.

[0154] See also Figure 10 , Figure 10 The flowchart corresponding to the live data processing in the embodiment of the present application includes:

[0155] Determining audience users: This is equivalent to determining multiple online objects in a specified group based on the target live broadcast.

[0156] Obtaining data to be played: This is equivalent to determining the queue of the first delivered object to be played and the information flow interface of the second delivered object.

[0157] Estimating whether users will watch live broadcasts: This is equivalent to determining the viewing probability based on the characteristic data of online objects, and then determining the target objects.

[0158] Counting the information to be played in the live broadcast room: This is equivalent to counting the number of target objects, the number of objects to be played, and the number of second objects to be delivered.

[0159] Sent to the host for display: Equivalent to the number of display target objects, the number of objects to be played, and the number of second objects to be delivered.

[0160] The following is an introduction to the training method of the neural network in this application. Figure 11a , Figure 11a The method flow chart of the neural network training method provided in the embodiment of the present application includes:

[0161] 1101, determining a training object associated with a target live broadcast;

[0162] 1102, acquiring feature data of a training object, the feature data being used to calculate a viewing probability, where the viewing probability indicates a probability of viewing a target live broadcast;

[0163] 1103, obtaining a predicted probability of the training object through a neural network based on the feature data;

[0164] 1104. Train the neural network according to a loss function, where the loss function indicates the similarity between the predicted probability and the expected probability, and the expected probability indicates the actual viewing situation of the target live broadcast by the training subject.

[0165] It is understandable that the embodiment of the present application uses a machine learning algorithm based on the object's behavioral characteristics to predict whether the object will watch the live broadcast.

[0166] In one possible implementation method, the feature data includes the ranking of the target live broadcast in the waiting-to-be-played queue of the training object, the exposure of the target live broadcast, and at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast and / or similar live broadcasts. Similar live broadcasts indicate live broadcasts of the same type as the target live broadcast.

[0167] Specifically, when the training object may include an active object of the target live broadcast, the characteristic data includes at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast. The training object may include a first delivered object, the to-be-played queue of the first delivered object includes at least one to-be-played live broadcast, the target live broadcast belongs to the to-be-played live broadcast, and the characteristic data of the first delivered object includes the order of the target live broadcast in the to-be-played queue. The training object may also include a second delivered object, the information flow interface of the second delivered object includes at least one exposed live broadcast, the target live broadcast belongs to the exposed live broadcast, and the characteristic data of the second delivered object includes at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the same type of live broadcast, and the same type of live broadcast indicates a live broadcast of the same type as the target live broadcast.

[0168] See also Figure 11b , Figure 11b The training flowchart of the neural network training method can be explained as the following main steps:

[0169] User action: Determine the live broadcast viewing behavior of the training subject;

[0170] Collecting features: Collecting corresponding feature data based on the above live viewing behavior;

[0171] Algorithm model: input feature data into the neural network model for prediction;

[0172] Prediction results: Obtain the prediction results of the neural network model, indicating the probability of the training subject watching the target live broadcast;

[0173] Execution result: Get the actual viewing situation of the target live broadcast by the training subject (including watching the live broadcast and not watching the live broadcast).

[0174] Optimize the model: Based on the similarity between the actual execution results and the predicted results, adjust the parameters in the neural network model so that the neural network model can output results that are closer to the actual results.

[0175] Specifically, we can use the logistic regression algorithm in the neural network algorithm to predict whether the subject will watch the live broadcast. The corresponding model is the logistic regression model. Figure 12 , Figure 12 This is a flow chart for training the logistic regression model provided in the embodiments of the present application.

[0176] Where X is the input feature data vector, and x_(index,u1) represents the feature data of user u1 with subscript index. For scenarios involving whether a target live stream is viewed, the target live stream's position in the viewing queue, the time interval since the user last viewed the target live stream, the user's average viewing time for the target live stream, the current network status, the user's viewing frequency within 7 days / 21 days, whether the user has tipped the target live stream within 7 days / 21 days, and whether the user has liked or commented on the target live stream within 7 days / 21 days.

[0177] Y is the result vector, which represents the result of the user watching the target live broadcast (watched or not watched). y_u1 represents the result of user u1 watching the target live broadcast, where watching is 1 and not watching is 0.

[0178] f(X) is the function that maps feature data to viewing outcomes. For a logistic regression model, f corresponds to the equation Y = g(W*X), where g is the Softmax function, which normalizes a real vector and converts it into a probability distribution. The output is mapped to the range (0, 1). W is the weight, which indicates the influence of each feature on the prediction outcome.

[0179] Since the training objects X and Y are known data, the weight W can be obtained according to Y = g(W*X), and the final output equation Y = g(W*X) is the model f.

[0180] In the prediction stage, the collected feature data of the training object is input through model f to obtain the result of predicting "whether the user watches the live broadcast", and the prediction result and the result of whether the training object actually watches the live broadcast are fed back to the model for model optimization.

[0181] The live data processing device in this application is described in detail below. Figure 13 . Figure 13 This is a schematic diagram of an embodiment of a live data processing device in an embodiment of the present application. The live data processing device 1300 includes:

[0182] A first determining module 1301 is configured to determine a designated group according to a target live broadcast, where the designated group includes a plurality of online objects associated with the target live broadcast;

[0183] The first acquisition module 1302 is further configured to acquire feature data of each online subject in a specified group, the feature data being used to calculate a viewing probability, where the viewing probability indicates a probability of viewing a target live broadcast;

[0184] A first prediction module 1303 is configured to obtain a viewing probability of an online object through a neural network based on the feature data;

[0185] Processing module 1304, configured to determine an online object with a viewing probability higher than a first preset value as a target object;

[0186] The output module 1305 is used to display the number of target objects through the terminal device corresponding to the target live broadcast.

[0187] The live broadcast data processing device provided in the embodiments of the present application first determines the group of online subjects associated with the target live broadcast, then obtains the characteristic data of each online subject in the group. This characteristic data is then used to predict the probability of each online subject watching the target live broadcast. Based on this probability, the target subjects who will watch the target live broadcast are determined, and finally, the number of target subjects is displayed on the terminal device corresponding to the target live broadcast. This method allows the host to determine the number of viewers who may enter the live broadcast room in real time during the live broadcast, thereby adjusting the live broadcast strategy accordingly.

[0188] In a possible implementation method, the multiple online objects include an active object of the target live broadcast, and the characteristic data of the active object includes at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast.

[0189] In this embodiment, the active user of the target live broadcast refers to a user who is active in the live broadcast room corresponding to the target live broadcast and is therefore likely to watch the target live broadcast. The active user has watched the target live broadcast before, so the probability of the active user watching the target live broadcast can be predicted based on the active user's viewing behavior of the target live broadcast.

[0190] In a possible implementation method, the plurality of online objects include a first delivered object, a to-be-played queue of the first delivered object includes at least one to-be-played live broadcast, and the target live broadcast belongs to the to-be-played live broadcast.

[0191] In this embodiment, if the target live broadcast is in the to-be-played list of the online object, then the target live broadcast is likely to be watched by the online object, so the online objects associated with the target live broadcast include the first delivered object.

[0192] In an optional case of a possible implementation method, the characteristic data of the first delivered object includes the order of the target live broadcast in the queue to be played.

[0193] In this embodiment, since the first delivery target includes the target live broadcast in the to-be-played list, the higher the ranking of the target live broadcast in the to-be-played list, the greater the likelihood of it being played. Therefore, the probability of the first delivery target watching the target live broadcast can be predicted based on the ranking of the target live broadcast in the to-be-played list.

[0194] In an alternative case of a possible implementation method,

[0195] The output module 1305 is also used to determine the order of the target live broadcast in the queue to be played based on the first delivered object; determine the first delivered object corresponding to the target live broadcast whose order is less than the second preset value as the object to be played; and display the number of objects to be played through the terminal device.

[0196] In this embodiment, since the target live broadcast is included in the queue of the first delivered object, the ranking of the target live broadcast in the queue can be obtained; the smaller the ranking, the higher the possibility of the target live broadcast being played, so the objects to be played can be further screened out from the first delivered object; displaying the number of objects to be played by the terminal device can help the anchor subjectively estimate the future number of viewers, and further assist the anchor in adjusting the live broadcast strategy.

[0197] In a possible implementation method, the plurality of online objects include a second delivered object, the information flow interface of the second delivered object includes at least one exposed live broadcast, and the target live broadcast belongs to the exposed live broadcast.

[0198] In this embodiment, in the scenario corresponding to the information flow interface, all types of live broadcasts generally go through the process of exposure, clicks, and viewing. Therefore, the exposed live broadcasts in the information flow interface have the possibility of being viewed. Therefore, the online objects including the target live broadcast in the information flow interface include the second delivered objects.

[0199] In an alternative case of a possible implementation method,

[0200] The characteristic data of the second delivery object includes at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the same type of live broadcast. The same type of live broadcast indicates the same type of live broadcast as the target live broadcast.

[0201] In this embodiment, the second recipient of the delivery needs to click on the corresponding live broadcast card to watch the target live broadcast. Therefore, the higher the second recipient's interest in the target live broadcast, the greater the probability of the target live broadcast being played. To determine the second recipient's interest in the target live broadcast, the second recipient's viewing behavior for similar live broadcasts can be used to determine the second recipient's interest in the target live broadcast.

[0202] In an alternative case of a possible implementation method,

[0203] The output module 1305 is further configured to display the number of the second delivered objects through the terminal device.

[0204] In this embodiment, the target live broadcast is exposed in the information flow interface of the second delivered object, so the number of the second delivered objects is displayed through the terminal device, which can help the anchor confirm the current traffic status.

[0205] In one possible implementation,

[0206] The output module 1305 is further configured to display the target object through a terminal device when the target object watches the target live broadcast.

[0207] In this embodiment, after the number of target objects is displayed to the terminal device corresponding to the target live broadcast, when the target objects actually watch the target live broadcast, the viewing status of the target objects can be displayed to the anchor to improve the anchor's trust in the estimated results of the live broadcast platform.

[0208] The following is a detailed description of the neural network training device in this application. Figure 14 . Figure 14 This is a schematic diagram of an embodiment of a neural network training device in an embodiment of the present application. The neural network training device 1400 includes:

[0209] The second determining module 1401 is used to determine a training object associated with the target live broadcast;

[0210] A second acquisition module 1402 is used to acquire feature data of the training object, the feature data is used to calculate the viewing probability, and the viewing probability indicates the probability of watching the target live broadcast;

[0211] The second prediction module 1403 is used to obtain the predicted probability of the training object through a neural network based on the feature data;

[0212] The training module 1404 is used to train the neural network according to the loss function, where the loss function indicates the similarity between the predicted probability and the expected probability, and the expected probability indicates the actual situation of the training subject watching the target live broadcast.

[0213] In one possible implementation method, the feature data includes the ranking of the target live broadcast in the waiting-to-be-played queue of the training object, the exposure of the target live broadcast, and at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast and / or similar live broadcasts. Similar live broadcasts indicate live broadcasts of the same type as the target live broadcast.

[0214] The neural network training device provided in the embodiment of the present application is similar to the above Figure 10The training method of the neural network in the corresponding embodiment corresponds to the specific description, please refer to the above, and will not be repeated here.

[0215] The embodiment of the present application also provides a computer device, including a memory and a processor, wherein the memory stores instructions, and when the instructions are executed on the processor, the computer device executes the following Figure 3 or Figure 5 The steps of the live data processing method in Figure 10 The steps of the neural network training method in .

[0216] The computer device may be Figure 2 The terminal device or server shown in the embodiment of the present application takes the computer device as an example to introduce the structure of the computing device. Figure 15 , Figure 15 This is a schematic diagram of a server structure provided in an embodiment of the present application. The server 300 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 322 (for example, one or more processors) and memory 332, and one or more storage media 330 (for example, one or more mass storage devices) for storing application programs 342 or data 344. Among them, the memory 332 and the storage medium 330 can be short-term storage or persistent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 322 can be configured to communicate with the storage medium 330 to execute a series of instruction operations in the storage medium 330 on the server 300.

[0217] The server 300 may also include one or more power supplies 326, one or more wired or wireless network interfaces 350, one or more input and output interfaces 358, and / or one or more operating systems 341, such as Windows Server 2003 or Windows Server 2003R. TM , Mac OS X TM , Unix TM ,Linux TM , FreeBSD TM etc.

[0218] The steps performed by the server in the above embodiment can be based on the Figure 15 The server structure shown.

[0219] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0220] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program that has a predetermined function and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0221] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0222] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0223] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0224] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0225] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A live broadcast data processing method, characterized in that: include: Determine a designated group according to the target live broadcast, the designated group including a plurality of online objects associated with the target live broadcast; Acquiring feature data of each of the online objects in the designated group, wherein the feature data is used to calculate a viewing probability, wherein the viewing probability indicates a probability of viewing the target live broadcast; Obtaining the viewing probability of the online object through a neural network according to the feature data; determining the online object with a viewing probability higher than a first preset value as a target object; The number of target objects is displayed through the terminal device corresponding to the target live broadcast.

2. The method according to claim 1, characterized in that The multiple online objects include active objects of the target live broadcast, and the feature data of the active objects include at least one of viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast.

3. The method according to claim 1, characterized in that The multiple online objects include a first delivered object, the to-be-played queue of the first delivered object includes at least one to-be-played live broadcast, and the target live broadcast belongs to the to-be-played live broadcast.

4. The method according to claim 3, characterized in that The characteristic data of the first delivered object includes the order of the target live broadcast in the queue to be played.

5. The method according to claim 3, characterized in that After determining the designated group according to the target live broadcast, the method further includes: Determining the order of the target live broadcast in the queue to be played based on the first delivered object; Determine the first delivered object corresponding to the target live broadcast whose rank is smaller than the second preset value as the object to be played; The number of the objects to be played is displayed via the terminal device.

6. The method according to claim 1, characterized in that The multiple online objects include a second delivered object, the information flow interface of the second delivered object includes at least one exposed live broadcast, and the target live broadcast belongs to the exposed live broadcast.

7. The method according to claim 6, characterized in that The characteristic data of the second delivery object includes at least one of viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to similar live broadcasts, and the similar live broadcasts indicate live broadcasts of the same type as the target live broadcast.

8. The method according to claim 6, characterized in that After determining the designated group according to the target live broadcast, the method further includes: The number of the second delivered objects is displayed through the terminal device.

9. The method according to claim 1, characterized in that After displaying the number of target objects through the terminal device corresponding to the target live broadcast, the method further includes: When the target object watches the target live broadcast, the target object is displayed through the terminal device.

10. A neural network training method, characterized in that: include: Determining a training object associated with the target live broadcast; Acquiring feature data of the training object, wherein the feature data is used to calculate a viewing probability, wherein the viewing probability indicates a probability of watching the target live broadcast; Obtaining a predicted probability of the training object through a neural network according to the feature data; The neural network is trained according to a loss function, where the loss function indicates a similarity between the predicted probability and an expected probability, where the expected probability indicates a situation in which the training subject actually watches the target live broadcast.

11. The method according to claim 10, characterized in that The characteristic data includes the ranking of the target live broadcast in the to-be-played queue of the training object, the exposure of the target live broadcast, and at least one of the viewing time, viewing frequency, network status, consumption behavior, and interactive behavior corresponding to the target live broadcast and / or similar live broadcasts. The similar live broadcast indicates a live broadcast of the same type as the target live broadcast.

12. A live broadcast data processing device, characterized in that: include: A first determining module is configured to determine a designated group according to a target live broadcast, wherein the designated group includes a plurality of online objects associated with the target live broadcast; The first acquisition module is further configured to acquire feature data of each online object in the designated group, wherein the feature data is used to calculate a viewing probability, where the viewing probability indicates a probability of viewing the target live broadcast; a first prediction module, configured to obtain the viewing probability of the online object through a neural network according to the feature data; A processing module, configured to determine the online object whose viewing probability is higher than a first preset value as a target object; An output module is used to display the number of target objects through a terminal device corresponding to the target live broadcast.

13. A neural network training device, characterized in that: include: A second determining module is used to determine a training object associated with the target live broadcast; A second acquisition module is used to acquire feature data of the training object, where the feature data is used to calculate a viewing probability, where the viewing probability indicates a probability of watching the target live broadcast; A second prediction module, configured to obtain a predicted probability of the training object through a neural network based on the feature data; A training module is used to train the neural network according to a loss function, wherein the loss function indicates the similarity between the predicted probability and the expected probability, and the expected probability indicates the situation in which the training subject actually watches the target live broadcast.

14. A computer device, characterized in that: include: memory and processor; The memory stores instructions, and when the instructions are run on the processor, the live data processing method as described in any one of claims 1 to 9 is executed, or the neural network training method as described in claim 10 or 11 is executed.

15. A computer-readable storage medium comprising instructions, which, when executed on a computer, enables the computer to execute the live data processing method according to any one of claims 1 to 9, or the neural network training method according to claim 10 or 11.

Images