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

Abstract

Embodiments of the present application disclose a data processing method, apparatus, device, and readable storage medium. The method includes: obtaining target video data; obtaining a target policy, and predicting whether the target video data belongs to a popular video based on the target policy, where the target policy includes at least one of a video hit policy and a video prediction policy; if the target video data belongs to a popular video, performing transcoding processing on the target video data to obtain transcoded video data; and when a request for obtaining the target video data is received, sending the transcoded video data to a terminal device. By using the embodiments of the present application, it is possible to determine whether the target video data is a popular video and improve data processing efficiency.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data processing method, device, equipment and readable storage medium. Background Art

[0002] Before playing video data, transcoding the video data can make the video quality clearer and reduce the size of the video file. However, the transcoding resources for video data are limited. It is necessary to find a small part of the video data with a large playback volume in the application, that is, popular videos. This part of popular video data can cover most video playbacks. Preprocessing this part of video data can improve the video playback experience of most users. At present, video data is generally judged manually to determine whether the video data is a popular video, and then the popular video is transcoded. This popular video judgment method is inefficient and will reduce data processing efficiency. Summary of the invention

[0003] The embodiments of the present application provide a data processing method, apparatus, device and readable storage medium, which can determine whether target video data is a popular video and improve data processing efficiency.

[0004] In a first aspect, the present application provides a data processing method, comprising:

[0005] Get target video data;

[0006] Obtaining a target strategy, and predicting whether the target video data is a hot video based on the target strategy, wherein the target strategy includes at least one of a video hit strategy and a video prediction strategy, wherein the video hit strategy is used to predict whether the target video data is a hot video based on feature data of a user, and the video prediction strategy is used to predict whether the target video data is a hot video based on a video prediction model;

[0007] If the target video data is a popular video, transcoding the target video data to obtain transcoded video data;

[0008] When an acquisition request for the target video data is received, the transcoded video data is sent to the terminal device.

[0009] In a second aspect, the present application provides a data processing device, characterized in that the device comprises:

[0010] A data acquisition unit, used for acquiring target video data;

[0011] A data prediction unit, configured to obtain a target policy and predict whether the target video data belongs to a popular video based on the target policy. The target policy includes at least one of a video hit policy and a video prediction policy. The video hit policy is used to predict whether the target video data belongs to a popular video according to the user's feature data, and the video prediction policy is used to predict whether the target video data belongs to a popular video according to a video prediction model.

[0012] A data transcoding unit, configured to transcode the target video data to obtain transcoded video data if the target video data belongs to a popular video.

[0013] A data sending unit, configured to send the transcoded video data to a terminal device when a request for obtaining the target video data is received.

[0014] Combined with the second aspect, in a possible implementation, the target policy includes a video hit policy. The video hit policy includes user attention and / or a whitelist. The user attention is used to reflect the degree of attention of the user, and the whitelist includes at least one user. The data prediction unit is specifically configured to:

[0015] Determine whether the user attention of the target user corresponding to the target video data is greater than an attention threshold and / or whether the target user hits the whitelist.

[0016] If the attention of the target user is greater than the attention threshold and / or the target user hits the whitelist, determine that the target video data belongs to a popular video.

[0017] If the attention of the target user is less than or equal to the attention threshold and / or the target user does not hit the whitelist, determine that the target video data does not belong to a popular video.

[0018] Combined with the second aspect, in a possible implementation, the target policy includes a video prediction policy. The video prediction policy includes multiple video prediction models, and the video prediction model is used to predict whether the target video data belongs to a popular video. The data prediction unit is specifically configured to:

[0019] Obtain service requirement data corresponding to the target video data, and determine a target video prediction model from the multiple video prediction models based on the service requirement data.

[0020] Use the target video prediction model to predict the target video data to obtain a video prediction result, where the video prediction result includes a video prediction score.

[0021] If the video prediction score is higher than a score threshold, determine that the target video data belongs to a popular video.

[0022] If the video prediction score is lower than or equal to the score threshold, it is determined that the target video data does not belong to a popular video.

[0023] Combined with the second aspect, in a possible implementation, the video prediction result includes a video prediction score; the data prediction unit is specifically configured to:

[0024] Based on the target video prediction model, predict the target video data to obtain the video score corresponding to the video features of the target video data and the user score corresponding to the user features respectively. The video features include the video play volume and the video interaction volume, and the user features include the user attention and the user play volume;

[0025] Obtain the video category features of the target video data, and use the target video prediction model to respectively determine the first weight corresponding to the video features and the second weight corresponding to the user features based on the video category features. The video category features are used to indicate the category to which the target video data belongs;

[0026] Use the target video prediction model to determine the video prediction score based on the video score, the first weight, the user score, and the second weight.

[0027] Combined with the second aspect, in a possible implementation, the data processing device further includes a model training unit, which is used to:

[0028] Obtain a plurality of sample video data and sample labels;

[0029] Use the initial video prediction model to predict each sample video data in the plurality of sample video data to obtain the sample prediction result of each sample video data;

[0030] Based on the sample prediction result of each sample video data and the sample label, determine the model deviation feature for the initial video prediction model;

[0031] Train the initial video prediction model based on the model deviation feature to obtain the target video prediction model.

[0032] Combined with the second aspect, in a possible implementation, the data processing device further includes a data comparison unit, which is used to:

[0033] Compare the target video data with the reference video data in the video database to determine the comparison result for the target video data. The comparison result is used to indicate whether the target video data is duplicate video data;

[0034] If the comparison result indicates that the target video data does not match each reference video data in the video database, perform the step of predicting whether the target video data belongs to a popular video based on the target strategy;

[0035] If the comparison result indicates that the target video data matches the target reference video data in the video database, determine whether the target video data belongs to a popular video based on the video tag corresponding to the target reference video data.

[0036] In combination with the second aspect, in a possible implementation manner, the data processing device further includes a policy adjustment unit, configured to:

[0037] Push the transcoded video data to at least one terminal device;

[0038] Detect, within a target time period, an operation instruction of the at least one terminal device for the transcoded video data, where the operation instruction is used to perform at least one of operations of playing, liking, favoriting, and forwarding the transcoded video data;

[0039] If the data operation amount corresponding to the operation instruction is less than a quantity threshold, adjust the target policy.

[0040] In a third aspect, the present application provides a computer device, including: a processor, a memory, and a network interface;

[0041] The above-mentioned processor is connected to the memory and the network interface. Among them, the network interface is used to provide a data communication function, the above-mentioned memory is used to store computer program code, and the above-mentioned processor is used to call the above-mentioned computer program code so that the computer device including the processor executes the above-mentioned data processing method.

[0042] In a fourth aspect, the present application provides a computer-readable storage medium, in which a computer program is stored, and the computer program is suitable for being loaded and executed by a processor so that a computer device with the processor executes the above-mentioned data processing method.

[0043] In a fifth aspect, the present application provides a computer program product or a computer program, where the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the data processing method provided in various alternative manners in the first aspect of the present application.

[0044] In the embodiments of the present application, by obtaining target video data and obtaining a target policy, it is predicted whether the target video data belongs to a popular video based on the target policy; if the target video data belongs to a popular video, transcoding processing is performed on the target video data to obtain transcoded video data; when a request for obtaining the target video data is received, the transcoded video data is sent to the terminal device. Since this method can automatically predict the obtained target video data, it can improve the prediction efficiency of popular videos, thereby enhancing the data processing efficiency. Further, the target policy includes at least one of a video hit policy and a video prediction policy. When the target policy includes a video hit policy, it is possible to quickly determine whether the target video data is a popular video based on the video hit policy, improving the prediction efficiency. When the target policy includes a video prediction policy, it is possible to predict whether the target video data is a popular video based on the video prediction policy, improving the data prediction accuracy. When the target policy includes a video hit policy and a video prediction policy, it is possible to preliminarily determine whether the target video data is a popular video based on the video hit policy, improving the data prediction accuracy. If it is preliminarily determined that the target video data does not belong to a popular video, the video prediction policy can be further used to perform a secondary determination on the target video data to determine whether the target video data belongs to a popular video, further improving the video prediction accuracy. In addition, by predicting popular videos and performing transcoding processing on the popular videos, the transcoded video data can be directly obtained later, which can save bandwidth costs and improve data transmission efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0046] Figure 1 is a network architecture diagram of a data processing system provided by an embodiment of the present application;

[0047] Figure 2 is a flowchart of a data processing method provided by an embodiment of the present application;

[0048] Figure 3 is a flowchart of another data processing method provided by an embodiment of the present application;

[0049] Figure 4 is a schematic diagram of the composition structure of a data processing device provided by an embodiment of the present application;

[0050] Figure 5 is a schematic diagram of the composition structure of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0051] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0052] The technical solution of the present application can be applied to the scenario of predicting whether video data belongs to popular video. By predicting whether the video data belongs to popular video, if the video data belongs to popular video, the video data can be transcoded to obtain transcoded video data, and then the transcoded video data can be sent to the terminal device, which can reduce the bit rate of the video data and the bandwidth cost.

[0053] Please refer to Figure 1 , Figure 1 which is a network architecture diagram of a data processing system provided by an embodiment of the present application. As Figure 1 shown, the computer device can interact with the terminal device, and the number of terminal devices can be one or at least two. For example, when the number of terminal devices is multiple, the terminal devices can include Figure 1 the terminal device 101a, terminal device 101b, and terminal device 101c in

[0054] In the embodiments of the present application, the computer device mentioned may refer to a server or a terminal device, or may also refer to a system composed of a server and a terminal device. The embodiments of the present application do not make any limitations thereto. The terminal device may be an electronic device, including but not limited to a mobile phone, a tablet computer, a laptop computer, a vehicle-mounted device, etc. The server mentioned above may be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or may also be a cloud server providing 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, vehicle-road collaboration, Content Delivery Network (CDN), and big data and artificial intelligence platforms. When the server is an independent physical server, the server can independently process data. When the server is composed of multiple physical servers, multiple physical servers can cooperate with each other to process data, etc.

[0055] Further, please refer to Figure 2 , Figure 2 which is a schematic flowchart of a data processing method provided by the embodiments of the present application; as Figure 2 shown, this data processing method can be applied to a computer device or a client. The client can be deployed on the computer device. This data processing method includes but is not limited to the following steps:

[0056] S101, obtain target video data.

[0057] In the embodiments of the present application, the computer device can obtain target video data from any terminal device or database. For example, the computer device can obtain target video data from a terminal device. For example, when the terminal device triggers an upload event for the target video data, the computer device detects the upload event for the target video data and thus obtains the target video data. For example, when the terminal device detects a trigger operation by the user on the upload button on the screen of the terminal device, the terminal device triggers an upload event for the target video data, and the computer device obtains the target video data. The target video data may refer to a continuous sequence of images. The target video data may be obtained by a video shooting device recording various elements in nature such as people, animals, plants, landscapes, etc., or may also be obtained by synthesizing multiple video data or image data. The embodiments of the present application do not make any limitations thereto.

[0058] Optionally, the computer device can also obtain the target video data from local storage, or obtain the target video data from a cloud database, or can also receive the target video data uploaded by any terminal device. The embodiments of the present application do not make any limitations in this regard. After obtaining the target video data, the computer device can predict the target video data to determine whether the target video data belongs to a popular video, so as to determine the processing method for the target video data. Among them, the popular video can be used to reflect the quantitative result of the popularity of the video. For example, the popular video can include, but is not limited to, videos with the user play volume greater than the play threshold, videos with the user favorite volume greater than the favorite threshold, videos with the user like volume greater than the like threshold, videos with the user share volume greater than the share threshold, and so on.

[0059] S102. Obtain a target policy, and predict whether the target video data belongs to a popular video based on the target policy.

[0060] In the embodiments of the present application, the computer device can obtain the target policy from any terminal device or the policy library, so as to predict whether the target video data belongs to a popular video based on the target policy. In a possible implementation manner, the computer device can obtain the target policy, so that for all video data, the target policy can be used to predict whether it belongs to a popular video. That is to say, the computer device can obtain a policy, so that for all video data, this policy can be used to determine whether it is a popular video. For example, if the target policy is used to indicate that when the click volume of the video data is greater than the click volume threshold, the video data is considered a popular video, then when any video data is obtained, if the click volume of the video data is greater than the click volume threshold, the video data is considered a popular video.

[0061] In another possible implementation manner, the computer device can obtain the target policy associated with the target video data, so as to predict whether the target video data belongs to a popular video based on the target policy associated with the target video data. For example, the computer device can obtain the target policy associated with the target video data from the policy library, and predict whether the target video data belongs to a popular video based on the target policy. The policy library can pre-store the association relationships between various policies and the categories of video data, such as which category of video data is applicable to which policy. When the target video data is obtained, the category of the target video data can be determined, so as to obtain the policy associated with the category of the target video data from the policy library as the target policy associated with the target video data. The policies associated with each category of video data can be different, and the policies can be specifically set in advance according to requirements.

[0062] In one embodiment, if the target policy includes a video hit policy, and the video hit policy may include user attention, where the user attention is used to reflect the degree of attention received by the user, then the method for predicting whether the target video data belongs to a popular video based on the target policy may include: determining whether the user attention of the target user corresponding to the target video data is greater than the attention threshold; if the attention of the target user is greater than the attention threshold, determining that the target video data belongs to a popular video; if the attention of the target user is less than or equal to the attention threshold, determining that the target video data does not belong to a popular video.

[0063] Among them, the target user may refer to the user of the terminal device that uploads the target video data. For example, it may refer to the user of the account registered on the terminal device. The user attention may, for example, refer to the number of fans of the user, that is, the number of users having a following relationship with the target user. The larger the number of fans of the user, the higher the user attention, and the greater the probability that the video published by the user becomes a popular video. If the target video data hits the video hit policy, it means that the target video data belongs to a popular video. If the target video data does not hit the video hit policy, it means that the target video data does not belong to a popular video.

[0064] It can be understood that for different categories of video data, the corresponding attention thresholds may be different. For example, for video data of certain popular categories (such as food category, funny category), the attention threshold may be a first value; for video data of certain unpopular categories (such as film and television category, history category), the attention threshold may be a second value; the first value is greater than the second value. That is to say, for different categories of video data, the corresponding attention thresholds are different. When the user's attention is greater than the attention threshold corresponding to the category, it can be determined that the target video data belongs to a popular video, otherwise it does not.

[0065] In one embodiment, if the video hit policy includes a whitelist, and the whitelist includes at least one user; then the method for predicting whether the target video data belongs to a popular video based on the target policy may include: determining whether the target user corresponding to the target video data hits the whitelist; if the target user hits the whitelist, determining that the target video data belongs to a popular video; if the target user does not hit the whitelist, determining that the target video data does not belong to a popular video.

[0066] Among them, the users in the whitelist can include, but are not limited to, authoritative users, star users, or expert users in various fields, and so on. It can be understood that for different categories of video data, the users in the corresponding whitelist can be different. For example, for video data of certain categories (such as food category, makeup category), the whitelist can include multiple food users, beauty users, and so on. For video data of other categories (such as film and television category), the whitelist can include multiple film and television commentary users, and so on. That is to say, for different categories of video data, their corresponding whitelists are different. When the target user hits the whitelist corresponding to the category associated with the target video data, it is determined that the target video data belongs to a popular video. When the target user does not hit the whitelist corresponding to the category associated with the target video data, it is determined that the target video data does not belong to a popular video.

[0067] In the embodiments of the present application, the video hit policy can be used to predict whether the target video data belongs to a popular video according to the feature data of the user. For example, the feature data of the user can include, but are not limited to, the degree of user attention and the whitelist, and the embodiments of the present application do not make limitations in this regard.

[0068] In one embodiment, if the video hit policy includes the user attention degree and the whitelist, the method for predicting whether the target video data belongs to a popular video based on the target policy may include: determining whether the user attention degree of the target user corresponding to the target video data is greater than the attention threshold and whether the target user hits the whitelist; if the attention degree of the target user is greater than the attention threshold and the target user hits the whitelist, it is determined that the target video data belongs to a popular video; if the attention degree of the target user is less than or equal to the attention threshold or the target user does not hit the whitelist, it is determined that the target video data does not belong to a popular video.

[0069] In the embodiments of the present application, by using the video hit policy to judge the target video data, it is equivalent to only needing to obtain the user attention degree of the target user corresponding to the target video data, determine whether the user attention degree of the target user is greater than the attention threshold, and whether the target user hits the whitelist, so as to determine whether the target video data belongs to a popular video. Popular videos can be quickly determined, the efficiency of judging popular videos can be improved, and then the transcoding process of popular videos can be carried out as soon as possible. When the terminal device needs to obtain the target video data subsequently, the data transmission efficiency can be improved.

[0070] In the embodiments of the present application, the video prediction policy can also be used to predict whether the target video data belongs to a popular video according to the target video prediction model. The target video prediction model can be used to predict the target video data and determine whether the target video data belongs to a popular video. The method for predicting whether the target video data belongs to a popular video according to the target video prediction model is described below:

[0071] In one embodiment, the target policy includes a video prediction policy. The video prediction policy may include video prediction network prediction, video prediction algorithm prediction, video prediction model prediction, or other video prediction rule predictions. In the embodiments of the present application, the video prediction policy including a video prediction model is taken as an example for illustration. The video prediction policy may include multiple video prediction models, and the video prediction models are used to predict whether the target video data belongs to a popular video. Then, the method for obtaining the target policy associated with the target video data and predicting whether the target video data belongs to a popular video based on the target policy may include:

[0072] Obtain the service demand data corresponding to the target video data, and determine the target video prediction model from multiple video prediction models based on the service demand data; use the target video prediction model to predict the target video data to obtain a video prediction result, where the video prediction result includes a video prediction score; if the video prediction score is higher than the score threshold, determine that the target video data belongs to a popular video; if the video prediction score is lower than or equal to the score threshold, determine that the target video data does not belong to a popular video.

[0073] Among them, the prediction result can be used to indicate that the target video data is a popular video or not. The video prediction policy can be used to determine whether the target video data belongs to a popular video based on the video prediction model. Since the video prediction policy includes multiple video prediction models, the aspects concerned by each video prediction model may be different. For example, one video prediction model pays more attention to the number of popular videos, another video prediction model pays more attention to the prediction delay of popular videos, and another video prediction model pays more attention to the playback coverage rate of popular videos. Then, in the specific application process, the corresponding video prediction model can be obtained according to the requirements to predict the target video data. The service demand data corresponding to the target video data may include, but is not limited to, the number of popular videos, the prediction delay of popular videos, the playback coverage rate of popular videos, and so on. That is, the computer device can obtain the service demand data sent by the terminal device or the service terminal, and thus select the corresponding video prediction model based on the service demand data to predict whether the target video data belongs to a popular video.

[0074] Further, use the target video prediction model to predict the target video data to obtain a video prediction result, and the video prediction result may include a video prediction score. Among them, the video prediction score can be used to reflect the possibility that the target video data belongs to a popular video. If the video prediction score is higher than the score threshold, it means that the possibility that the target video data belongs to a popular video is greater, and it is determined that the target video data belongs to a popular video. If the video prediction score is lower than or equal to the score threshold, it means that the possibility that the target video data belongs to a popular video is smaller, and it is determined that the target video data does not belong to a popular video.

[0075] Optionally, predicting the target video data using the target video prediction model to obtain a video prediction score may refer to predicting scores of multiple features corresponding to the target video data using the target video data, and determining the video prediction score based on the scores of each feature using the target video data. The multiple features may include, but are not limited to, video features and user features. Then, the video score corresponding to the video features of the target video data and the user score corresponding to the user features may be predicted respectively, and the video prediction score may be determined based on the video score corresponding to the video features and the user score corresponding to the user features. Among them, the video features may include, but are not limited to, video play volume and video interaction volume, and the user features may include, but are not limited to, user attention and user play volume. The video play volume may refer to the number of times the video in the target video data is played within the first time period, and the video interaction volume may include, but is not limited to, the number of likes for the video being liked, video forwarding volume, video collection volume, and so on. The user play volume may refer to the number of times the same user plays the target video data. Optionally, the video features may further include features such as video duration and video vertical category, and then whether the target video data belongs to a popular video may be determined in combination with features such as video duration and video vertical category later.

[0076] Among them, the video vertical category may refer to the category of the video, for example, it may include, but is not limited to, film and television category, food category, scenery category, and so on. Different video vertical categories may have different play volumes for the video data to be a popular video. For example, if the play volume of a film and television category video reaches the first play threshold, it may be considered that the video data is a popular video; if the play volume of a food category video reaches the second play threshold, it may be considered that the video data is a popular video. The first play threshold may be less than the second play threshold. By combining the video vertical category to predict whether the target video data is a popular video, the accuracy of popular video prediction can be improved.

[0077] In one embodiment, after determining the target video prediction model by obtaining the business requirement data of the target video data, the target video prediction model can be used to predict the target video data to obtain a video prediction result. Specifically, the computer device can predict the target video data based on the target video prediction model to respectively obtain a video score corresponding to the video features of the target video data and a user score corresponding to the user features. The video features include the video playback volume and the video interaction volume, and the user features include the user attention and the user playback volume; obtain the video category features of the target video data, and use the target video data to respectively determine a first weight corresponding to the video features and a second weight corresponding to the user features based on the video category features. The video category features are used to indicate the category to which the target video data belongs; furthermore, use the target video data to determine a video prediction score based on the video score, the first weight, the user score, and the second weight. That is to say, when predicting the target video data, a video score corresponding to the video features of the target video data and a user score corresponding to the user features are respectively obtained; obtain the video category features of the target video data, and respectively determine a first weight corresponding to the video features and a second weight corresponding to the user features based on the video category features; based on the video score, the first weight, the user score, and the second weight, determine the video prediction score. These processes can all be executed in the target video prediction model, that is, they can be executed internally by the target video prediction model. That is, only by inputting the target video data into the target video prediction model, the above processing of the target video data can be performed by the target video prediction model to output the video prediction score.

[0078] Optionally, the target video prediction model can include, for example, but not limited to, a convolutional neural network model (Convolutional Neural Network, CNN), a deep neural network model (Deep Neural Networks, DNN), a recurrent neural network model (Recurrent neural network, RNN), a residual network model (ResNet), and so on.

[0079] Among them, the video category features can be used to reflect the degree of attention to the features in all aspects. For example, in some scenarios (such as when the video interaction volume is the most concerned feature), the attention to the video features is greater than that of the user features, so the first weight is greater than the second weight. In other scenarios (such as when the user attention is the most concerned feature), the attention to the video features is less than that of the user features, so the first weight is less than the second weight. Optionally, the first weight and the second weight can be preset, so that the first weight and the second weight can be obtained during the processing of the target video prediction model. Furthermore, the target video prediction model can determine the video prediction score by combining the weights and the scores corresponding to various features in the video data, which can improve the accuracy of video prediction and thus improve the accuracy of popular video judgment.

[0080] In an alternative implementation, the video scores corresponding to the video features of the target video data may specifically include: if the number of times the video of the target video data is played within the first time period is greater than the number threshold, then determine the score corresponding to the video play volume as the first play score; if the number of times the video of the target video data is played within the first time period is less than or equal to the number threshold, then determine the score corresponding to the video play volume as the second play score, and the first play score is higher than the second play score. If the number of likes for the target video data is greater than the like threshold within the first time period, then determine the score corresponding to the video like volume as the first like score; if the number of likes for the target video data is less than or equal to the like threshold within the first time period, then determine the score corresponding to the video like volume as the second like score. If the number of forwards for the target video data is greater than the forward threshold within the first time period, then determine the score corresponding to the video forward volume as the first forward score; if the number of forwards for the target video data is less than or equal to the forward threshold within the first time period, then determine the score corresponding to the video forward volume as the second forward score. If the number of collections for the target video data is greater than the collection threshold within the first time period, then determine the score corresponding to the video collection volume as the first collection score; if the number of collections for the target video data is less than or equal to the collection threshold within the first time period, then determine the score corresponding to the video collection volume as the second collection score, and so on.

[0081] In one embodiment, before using the target video prediction model to predict the target video data, the target video prediction model may also be pre-trained so that the trained target video prediction model has the ability to predict the target video data to obtain a video prediction result. Specifically, multiple sample video data and sample labels can be obtained; the initial video prediction model is used to predict each sample video data among the multiple sample video data to obtain a sample prediction result for each sample video data; based on the sample prediction result of each sample video data and the sample label, a model deviation feature for the initial video prediction model is determined; and the initial video prediction model is trained based on the model deviation feature to obtain the target video prediction model.

[0082] Among them, the sample label can refer to the true value of the sample. For example, the sample label can be obtained by manually annotating the sample video data, and the sample label can be used to indicate whether the sample video data is a popular video. One sample video data corresponds to one sample label. The sample prediction result obtained by the initial video prediction model can refer to the network output value. Then, the initial video prediction model can be trained based on the true value of the sample and the network output value, so that the network output value approaches the true value of the sample as much as possible. The purpose of training the initial video prediction model is to make the true value of the sample and the network output value as consistent as possible. When the model deviation feature of the initial video prediction model (i.e., the first loss function of the initial video prediction model) is greater than the first feature threshold, it can indicate that the true value of the sample and the network output value are inconsistent, that is, the accuracy of the initial video prediction model is low. The model parameters in the initial video prediction model can be continuously adjusted to reduce the model deviation feature in the initial video prediction model. When the model deviation feature of the initial video prediction model is less than or equal to the first feature threshold, it can indicate that the true value of the sample and the network output value are consistent. It can be considered that the training of the initial video prediction model meets the requirements, and then the initial video prediction model at this time can be saved as the target video prediction model. Subsequently, the target video prediction model can be used to predict the target video data to obtain the video prediction result.

[0083] It can be understood that in the embodiments of the present application, the training is performed on the target video prediction model among multiple video prediction models. For other video prediction models among multiple video prediction models, this training method can also be referred to for training, so as to obtain the trained video prediction models. The embodiments of the present application do not describe this in detail.

[0084] Since when training the model, the target video prediction model can pay attention to various indicators such as the playback volume, like volume, forward volume, favorite volume, user attention, total playback volume, and video vertical category of the target video data, when the target video prediction model is used to predict the target video data subsequently, the target video prediction model can determine the scores of the target video data for these indicators. Then, the target video prediction model can further combine the weights corresponding to each indicator and the scores of each indicator to determine the prediction score for the target video data, so as to determine the prediction result for the target video data, and further determine whether the target video data belongs to a popular video.

[0085] In the embodiments of the present application, since a large number of sample video data are used to train the initial video prediction model during the training of the initial video prediction model, the initial video prediction model can learn more detailed information in the sample video data, such as the number of video likes, the number of video plays, the number of video forwards, the number of video collections, the video vertical category, the video duration, user attention, user plays, etc. Therefore, the sample prediction result corresponding to the sample video data predicted based on these detailed information can more accurately reflect whether the sample video data is a popular video. Since a large number of sample video data are used to train the initial video prediction model, the accuracy of the model can be improved, and the accuracy of video prediction can be improved when using the target video prediction model for popular video prediction subsequently.

[0086] In an optional implementation manner, after multiple target video prediction models are trained, the video play coverage rate of the popular videos predicted based on the target video prediction models can be detected in real time within a second time period to adjust the target video prediction model, so that the target video prediction model can more accurately predict popular videos. Alternatively, the target video prediction model can be adjusted based on the prediction delay of the target video data, so that the target video prediction model can predict popular videos more quickly, that is, shorten the time interval between the upload time of the target video data and the prediction time for the target video data, improve the popular video prediction efficiency, and can predict popular videos as soon as possible and perform transcoding, and send the transcoded video to the user for playing to improve the video playing efficiency. Moreover, when using the target video prediction model for prediction, the target video prediction model can be selected according to specific business requirements, and a video prediction model that better meets the business requirements can be selected for prediction to improve the prediction accuracy.

[0087] Optionally, the computer device can execute the video hit policy and the video prediction policy in parallel to respectively determine whether the target video data belongs to a popular video. When any one of the policies determines that the target video data belongs to a popular video, the target video data is transcoded to improve the data processing efficiency.

[0088] Optionally, the computer device may also first use a video hit strategy to judge the target video data to determine whether the target video data belongs to a hot video. If the target video data does not belong to a hot video, the video prediction strategy is further used to secondarily predict whether the target video data belongs to a hot video. By first using a video hit strategy to make a preliminary judgment on the target video data to determine whether the target video data belongs to a hot video, since this judgment method is more convenient, the efficiency of hot video judgment can be improved. Further, if it is determined through preliminary judgment that the target video data does not belong to a hot video, the target video data can also be secondarily predicted to determine whether the target video data belongs to a hot video, which can improve the accuracy of hot video prediction. Moreover, by judging whether the target video belongs to a hot video, only the hot video needs to be transcoded later, and the hot video can cover most of the video playback on the target platform. By transcoding this part of the hot video, the user's playback experience can be improved.

[0089] It can be understood that when any video data is obtained, by using the target strategy to determine whether the video data is a popular video, it is possible to quickly determine whether the video data is a popular video, thereby quickly determining whether to transcode the video data, which can reduce the delay in popular video prediction and improve the efficiency of popular video judgment.

[0090] S103: If the target video data is a popular video, transcoding is performed on the target video data to obtain transcoded video data.

[0091] In an embodiment of the present application, if the target video data belongs to a popular video, the target video data can be finely encoded to obtain transcoded video data. Fine encoding processing may refer to transcoding the target video data. Video transcoding may refer to encoding a video to compress the video data. Specifically, the transcoding process may refer to converting the target video data from a first bit rate to a second bit rate, and the first bit rate is greater than the second bit rate. That is to say, before the terminal device initiates an access request for the target video data, the target video data is converted into transcoded video data with a smaller bit rate by performing bit rate conversion on the target video data in advance. As the actual playback volume of the target video data increases, CDN (Content Delivery Network) bandwidth can be saved during resource transmission, thereby reducing communication overhead during resource transmission.

[0092] By determining whether the target video data is a popular video, if the target video data is a popular video, the target video data can be transcoded as soon as possible to obtain transcoded video data. When a request for obtaining the target video data is subsequently received from a terminal device, the transcoded video data can be sent to the corresponding terminal device, so that users can have a better experience sooner, and at the same time, more CDN bandwidth can be saved and resource overhead can be saved.

[0093] S104: When receiving a request for obtaining target video data, the transcoded video data is sent to the terminal device.

[0094] In an embodiment of the present application, since the target video data is transcoded to obtain transcoded video data, when a request to obtain the target video data is received, the transcoded video data can be sent to the terminal device without sending the untranscoded target video data to the terminal device. This can make the video quality clearer and the video file smaller, thereby improving the user's playback experience, and can also save CDN bandwidth costs and resource overhead.

[0095] In the embodiment of the present application, when receiving the target video data uploaded by the terminal device, the computer device can pre-judge the target video data as a popular video, thereby determining whether to transcode the target video data. If the target video data is transcoded to obtain transcoded video data, when receiving a request for obtaining the target video data from any terminal device, the transcoded video data can be sent to the corresponding terminal device, thereby playing the transcoded video data on the terminal device.

[0096] In an embodiment of the present application, target video data is obtained, and a target policy is obtained. Based on the target policy, it is predicted whether the target video data belongs to a popular video. If the target video data belongs to a popular video, the target video data is transcoded to obtain transcoded video data. When a request for obtaining the target video data is received, the transcoded video data is sent to a terminal device. Since this method can automatically predict the obtained target video data, the prediction efficiency of popular videos can be improved, thereby enhancing the data processing efficiency. Further, the target policy includes at least one of a video hit policy and a video prediction policy. When the target policy includes a video hit policy, it is possible to quickly determine whether the target video data is a popular video based on the video hit policy, improving the prediction efficiency. When the target policy includes a video prediction policy, it is possible to predict whether the target video data is a popular video based on the video prediction policy, improving the data prediction accuracy. When the target policy includes a video hit policy and a video prediction policy, it is possible to preliminarily determine whether the target video data is a popular video based on the video hit policy, improving the data prediction accuracy. If it is preliminarily determined that the target video data does not belong to a popular video, the video prediction policy can be further used to perform a secondary determination on the target video data to determine whether the target video data belongs to a popular video, further improving the video prediction accuracy. In addition, by predicting popular videos and transcoding the popular videos, the transcoded video data can be directly obtained subsequently, which can save bandwidth costs and improve data transmission efficiency.

[0097] Further, please refer to Figure 3 , Figure 3 which is a schematic flowchart of another data processing method provided by an embodiment of the present application. As Figure 3 shown, this data processing method can be applied to a computer device or a client, and the client can be deployed on the computer device. This data processing method includes but is not limited to the following steps:

[0098] S201, obtain target video data.

[0099] In an alternative implementation, when the target video data is obtained, it is also possible to determine whether the target video data is duplicate video data, thereby improving the data processing efficiency. Specifically, the target video data can be compared with the reference video data in the video database to determine the comparison result for the target video data. The comparison result is used to indicate whether the target video data is duplicate video data. If the comparison result indicates that the target video data matches the target reference video data in the video database, then based on the video tag corresponding to the target reference video data, it is determined whether the target video data belongs to a popular video. If the comparison result indicates that the target video data does not match each reference video data in the video database, then based on the target policy, it is predicted whether the target video data belongs to a popular video. For example, based on at least one of user attention, whitelist, and target video prediction model, it is predicted whether the target video data belongs to a popular video.

[0100] Among them, the video database can store multiple reference video data and the video tags of each reference video data. The video tags of the reference video data can be obtained by pre-predicting multiple reference video data in the video database. The video tags can be used to indicate whether the reference video data belongs to a popular video. By comparing the target video data with multiple reference video data, it can be determined whether the reference video data matches the target video data. If there is a target reference video data that matches the target video data, then the video tag of the target reference video data can be determined as the video tag of the target video data. For example, if the video tag of the target reference video data indicates that the target reference video data does not belong to a popular video, then the target video data does not belong to a popular video, and there is no need to perform subsequent prediction and transcoding processing on the target video data, which can improve the data processing efficiency. If the video tag of the target reference video data indicates that the target reference video data belongs to a popular video, then the target video data belongs to a popular video, and there is no need to perform subsequent prediction on the target video data and it can be directly transcoded, which improves the data processing efficiency.

[0101] Optionally, it is also possible to obtain the target reference transcoding data corresponding to the target reference video data from the video database, and determine the target reference transcoding data as the transcoded video data corresponding to the target video data, so there is no need to perform transcoding processing on the target video data again, which improves the data processing efficiency.

[0102] S202, Determine whether the target video data belongs to a popular video based on user attention.

[0103] S203, Determine whether the target video data belongs to a popular video based on the whitelist.

[0104] In the embodiments of the present application, a computer device may trigger a service based on popularity to predict target video data and determine whether the target video data belongs to a popular video. Among them, the popularity-triggered service may include, but is not limited to, prediction methods such as user attention, whitelist, and target video prediction model. If the target video data does not belong to a popular video, the target video data may not be transcoded, or a relatively simple transcoding method may be used to transcode the target video data to reduce the size of the target video data to a certain extent and reduce the transcoding complexity. Alternatively, other simple data processing methods may be used to process the target video data, which is not limited in the embodiments of the present application. Further, the implementation manners in steps S201 to S203 may refer to the implementation manners in the above steps S101 to S102, which will not be elaborated here.

[0105] S204. Use the target video prediction model to predict the target video data and determine whether the target video data belongs to a popular video.

[0106] In an alternative implementation manner, the computer device may obtain click data, which may include the target video data, and directly use the target video prediction model to predict the target video data to determine whether the target video data is a popular video. The target video data may include the target video image, the upload time of the target video, the time for obtaining the prediction result of the target video, and so on. Thus, the prediction delay of the target video data can be determined according to the upload time of the target video. In the embodiments of the present application, the specific method for using the target video prediction model to predict the target video data may refer to the implementation manner in the above step S102, which will not be elaborated here.

[0107] In the embodiments of the present application, since transcoding video data takes a long time, transcoding all video data will result in low video transcoding efficiency, and many video data in the target platform are duplicate video data. Therefore, by predicting each video data in the target platform to determine whether the video data is a popular video and transcoding the popular video data in the target platform, the amount of video transcoding can be reduced, and the user experience can be improved and the video transcoding cost can be reduced.

[0108] S205. If the target video data belongs to a popular video, deduplicate the target video data.

[0109] In the embodiments of the present application, duplicate removal of popular videos can be performed in real time. For example, duplicate removal can be performed on popular videos within a third time period. For example, within the third time period after a certain video data is determined to be a popular video, if the popular video is obtained again, prediction of the popular video is not performed, thereby improving data processing efficiency. The third time period can be, for example, 12 hours, 24 hours, 48 hours, etc., and the embodiments of the present application do not limit this. By performing duplicate removal on the target video data, the number of popular video predictions can be reduced, and data processing efficiency can be improved.

[0110] S206. Perform transcoding processing on the duplicate-removed target video data to obtain transcoded video data.

[0111] In the embodiments of the present application, after determining that the target video data is a popular video, a transcoding workflow can be triggered, and the corresponding transcoding workflow is used to perform transcoding on the target video data to obtain transcoded video data. Subsequently, when any terminal device needs to obtain the target video data, the transcoded video data can be sent to the terminal device, thereby improving video transmission efficiency.

[0112] In a possible implementation manner, different workflows can also be selected in combination with the acquisition situation of the video data and the prediction situation of the video data, so as to select the transcoding strategy corresponding to the workflow to perform transcoding on the video data. For example, it can include a first workflow, a second workflow, and a third workflow. Any one of the three workflows can be used to perform transcoding on the video data, but the transcoding conditions and / or transcoding results of the three workflows can be different. For example, the first workflow can be a workflow for preprocessing all video data, the second workflow can be a workflow for transcoding popular videos, and the third workflow can be a workflow for transcoding specified video data. The specified video data can include popular video data or any data other than popular video data.

[0113] For example, when the target video data is video data uploaded by a user, for example, after the user selects video data from the local repository of the terminal device and uploads it, when it is detected that the user triggers a note publishing operation, the first workflow can be started to preprocess the target video data to improve the upload efficiency of the target video data. Subsequently, the preprocessed target video data can be predicted to determine whether it is popular video data, and corresponding processing can be performed accordingly. Or when the target video data is video data stored in a database associated with the platform, the first workflow may not be started. Optionally, if it is determined through subsequent prediction of the target video data that the target video data is a popular video, the second workflow can also be started to perform transcoding processing on the target video data to obtain transcoded video data. Or when the target video data is specified video data, the third workflow can be started to process the target video data to obtain transcoded video data.

[0114] In an embodiment of the present application, preprocessing video data by starting the first workflow can quickly transcode the video data, thereby improving the publishing efficiency of the video data. Transcoding the video data through the second workflow or the third workflow can improve the clarity of video transcoding and reduce the file size, thereby saving bandwidth costs. Different transcoding workflows can be flexibly configured according to different business requirements. For example, in the case of a large amount of video data, the first workflow can be used for quick processing, which can improve the video distribution efficiency. For popular videos, the second workflow can be adopted or the third workflow can be specified for transcoding the video data, which can make the transcoded video clearer, the file smaller, and the bandwidth reduced, so as to achieve targeted data processing and improve the user experience. Since when it is determined that the target video data is a popular video, transcoding the target video data can reduce the video bit rate, so when the user plays the transcoded target video data, the video playback efficiency can be improved, the user waiting time can be reduced, and the user experience can be improved.

[0115] In one implementation, when an upload event for the target video data is detected, the target video data can be obtained, and the first workflow can be started to preprocess the target video data; the target policy can be obtained, and based on the target policy, it can be predicted whether the preprocessed target video data belongs to a popular video. If the preprocessed target video data belongs to a popular video, the second workflow can be started to transcode the popular video to obtain transcoded video data; when a request for obtaining the target video data is received, the transcoded video data is sent to the terminal device. Alternatively, if it is determined that the preprocessed target video data is specified video data, the third workflow can be used to transcode the preprocessed target video data to obtain transcoded video data; when a request for obtaining the target video data is received, the transcoded video data is sent to the terminal device.

[0116] S207, push the transcoded video data to at least one terminal device, detect the operation instructions of at least one terminal device for the transcoded video data within the target time period, and adjust the target policy based on the operation instructions.

[0117] In the embodiment of the present application, since the transcoded video data is obtained by transcoding the target video data, the transcoded video data can be pushed to the terminal devices used by multiple registered users in the target platform, so as to determine whether the target policy needs to be adjusted according to the operation instructions of the terminal devices for the transcoded video data.

[0118] Specifically, transcoded video data can be pushed to at least one terminal device; operation instructions of at least one terminal device for the transcoded video data are detected within a target time period, and the operation instructions are used for performing at least one of playing, liking, favoriting, and forwarding operations on the transcoded video data; if the data operation volume corresponding to the operation instructions is less than a quantity threshold, the target policy is adjusted.

[0119] For example, when the data operation volume corresponding to the operation instructions of at least one terminal device for the transcoded video data within the target time period is less than the quantity threshold, it can be considered that the target video data is not a popular video. Then, it can be indicated that it is incorrect to determine the target video data as a popular video using the target policy, and the target policy can be adjusted. For example, when the click-through rate of a video uploaded by a certain user in the whitelist is lower than the target threshold, the user is removed from the whitelist. That is, when the video uploaded by this user is subsequently obtained, it is initially determined that the video is not a popular video through the video hit policy, and it can be determined that the video data is not a popular video. Or the video prediction strategy such as the target video prediction model can be further used to judge the video data to further determine whether the video is a popular video, and so on.

[0120] In a possible implementation manner, when pushing videos, user tags can also be obtained, and popular videos are pushed to terminal devices that match the user tags to achieve targeted video pushing, improve the click-through rate of users, and thus improve the user experience.

[0121] In the embodiments of this application, through experimental statistics, it is obtained that the prediction accuracy rate of using this video prediction method is as high as 95%, and the playback coverage rate of popular videos reaches 90%. The accuracy rate and coverage rate of the transcoding heat trigger service based on the video prediction model are high. Only a small part of popular video data needs to be transcoded to cover 90% of video playback. And the heat trigger service is flexible and reliable, and the trigger policy and transcoding workflow for popular videos can be flexibly configured according to business needs. Moreover, by detecting data such as the number of popular videos, prediction latency, and playback coverage rate of popular videos in real time, the video data prediction model and video hit policy can be adjusted according to data performance and business needs to improve the prediction accuracy of popular videos.

[0122] In an embodiment of the present application, target video data is obtained, and a target policy is obtained. Based on the target policy, it is predicted whether the target video data belongs to a popular video. If the target video data belongs to a popular video, transcoding processing is performed on the target video data to obtain transcoded video data. When a request for obtaining the target video data is received, the transcoded video data is sent to a terminal device. Since this method can automatically predict the obtained target video data, the prediction efficiency of popular videos can be improved, and thus the data processing efficiency can be enhanced. Further, the target policy includes at least one of a video hit policy and a video prediction policy. When the target policy includes the video hit policy, it is possible to quickly determine whether the target video data is a popular video based on the video hit policy, improving the prediction efficiency. When the target policy includes the video prediction policy, it is possible to predict whether the target video data is a popular video based on the video prediction policy, improving the data prediction accuracy. When the target policy includes both the video hit policy and the video prediction policy, it is possible to preliminarily determine whether the target video data is a popular video based on the video hit policy, improving the data prediction accuracy. If it is preliminarily determined that the target video data does not belong to a popular video, the video prediction policy can be further used to perform a secondary determination on the target video data to determine whether the target video data belongs to a popular video, further improving the video prediction accuracy. In addition, by predicting popular videos and performing transcoding processing on the popular videos, the transcoded video data can be directly obtained later, which can save bandwidth costs and improve data transmission efficiency.

[0123] The method of the embodiment of the present application is introduced above. Next, the device of the embodiment of the present application is introduced.

[0124] See Figure 4 , Figure 4 is a schematic structural diagram of a composition of a data processing device provided by an embodiment of the present application. The above data processing device can be deployed on a computer device; the data processing device can be used to execute corresponding steps in the data processing method provided by the embodiment of the present application. Optionally, the data processing device can also be a client. The data processing device 40 includes:

[0125] A data acquisition unit 401, configured to acquire target video data;

[0126] A data prediction unit 402, configured to acquire a target policy and predict whether the target video data belongs to a popular video based on the target policy. The target policy includes at least one of a video hit policy and a video prediction policy; the video hit policy is used to predict whether the target video data belongs to a popular video according to the user's characteristic data, and the video prediction policy is used to predict whether the target video data belongs to a popular video according to a video prediction model;

[0127] A data transcoding unit 403, configured to perform transcoding processing on the target video data to obtain transcoded video data if the target video data belongs to a popular video;

[0128] A data sending unit 404, configured to send the transcoded video data to a terminal device when a request for obtaining the target video data is received.

[0129] Optionally, the target policy includes a video hit policy, the video hit policy includes user attention and / or a whitelist, the user attention is used to reflect the degree of attention of a user, and the whitelist includes at least one user; the data prediction unit 402 is specifically configured to:

[0130] Determine whether the user attention of the target user corresponding to the target video data is greater than an attention threshold and / or whether the target user hits the whitelist;

[0131] If the attention of the target user is greater than the attention threshold and / or the target user hits the whitelist, determine that the target video data belongs to a popular video;

[0132] If the attention of the target user is less than or equal to the attention threshold and / or the target user does not hit the whitelist, determine that the target video data does not belong to a popular video.

[0133] Optionally, the target policy includes a video prediction policy, the video prediction policy includes multiple video prediction models, and the video prediction model is used to predict whether the target video data belongs to a popular video; the data prediction unit 402 is specifically configured to:

[0134] Obtain service requirement data corresponding to the target video data, and determine a target video prediction model from the multiple video prediction models based on the service requirement data;

[0135] Use the target video prediction model to predict the target video data to obtain a video prediction result, where the video prediction result includes a video prediction score;

[0136] If the video prediction score is higher than a score threshold, determine that the target video data belongs to a popular video;

[0137] If the video prediction score is lower than or equal to the score threshold, determine that the target video data does not belong to a popular video.

[0138] Optionally, the video prediction result includes a video prediction score; the data prediction unit 402 is specifically configured to:

[0139] Predict the target video data based on the target video prediction model to obtain the video score corresponding to the video features of the target video data and the user score corresponding to the user features respectively. The video features include the video play volume and the video interaction volume, and the user features include the user attention and the user play volume;

[0140] Obtain the video category features of the target video data, and use the target video prediction model to respectively determine the first weight corresponding to the video features and the second weight corresponding to the user features based on the video category features. The video category features are used to indicate the category to which the target video data belongs;

[0141] Use the target video prediction model to determine the video prediction score based on the video score, the first weight, the user score, and the second weight.

[0142] Optionally, the data processing device 40 further includes a model training unit 405 for:

[0143] Obtain a plurality of sample video data and sample labels;

[0144] Use the initial video prediction model to predict each sample video data in the plurality of sample video data to obtain the sample prediction result of each sample video data;

[0145] Based on the sample prediction result of each sample video data and the sample label, determine the model deviation feature for the initial video prediction model;

[0146] Train the initial video prediction model based on the model deviation feature to obtain the target video prediction model.

[0147] Optionally, the data processing device 40 further includes a data comparison unit 406 for:

[0148] Compare the target video data with the reference video data in the video database to determine the comparison result for the target video data. The comparison result is used to indicate whether the target video data is duplicate video data;

[0149] If the comparison result indicates that the target video data does not match each reference video data in the video database, then perform the step of predicting whether the target video data belongs to a popular video based on the target strategy;

[0150] If the comparison result indicates that the target video data matches the target reference video data in the video database, then determine whether the target video data belongs to a popular video based on the video label corresponding to the target reference video data.

[0151] Optionally, the data processing device 40 further includes a strategy adjustment unit 407 for:

[0152] Push the transcoded video data to at least one terminal device;

[0153] Detect an operation instruction of the at least one terminal device for the transcoded video data within a target time period, where the operation instruction is used for at least one of playing, liking, favoriting, and forwarding the transcoded video data;

[0154] If the data operation amount corresponding to the operation instruction is less than a quantity threshold, adjust the target policy.

[0155] It should be noted that Figure 4 For the content not mentioned in the corresponding embodiments, reference can be made to the description of the method embodiments, which will not be elaborated here.

[0156] In the embodiments of the present application, target video data is obtained, and a target policy is obtained. Based on the target policy, it is predicted whether the target video data belongs to a popular video; if the target video data belongs to a popular video, transcoding processing is performed on the target video data to obtain transcoded video data; when a request for obtaining the target video data is received, the transcoded video data is sent to the terminal device. Since this method can automatically predict the obtained target video data, it can improve the prediction efficiency of popular videos, thereby improving the data processing efficiency. Further, the target policy includes at least one of a video hit policy and a video prediction policy. When the target policy includes a video hit policy, it is possible to quickly determine whether the target video data is a popular video based on the video hit policy, improving the prediction efficiency. When the target policy includes a video prediction policy, it is possible to predict whether the target video data is a popular video based on the video prediction policy, improving the data prediction accuracy. When the target policy includes a video hit policy and a video prediction policy, it is possible to preliminarily determine whether the target video data is a popular video based on the video hit policy, improving the data prediction accuracy. If it is preliminarily determined that the target video data does not belong to a popular video, the video prediction policy can be further used to perform a secondary determination on the target video data to determine whether the target video data belongs to a popular video, further improving the video prediction accuracy. In addition, by predicting popular videos and performing transcoding processing on the popular videos, the transcoded video data can be directly obtained later, which can save bandwidth and improve the data transmission efficiency.

[0157] See Figure 5 , Figure 5 is a schematic structural diagram of a computer device provided by the embodiments of the present application. As Figure 5As shown in the figure, the above computer device 50 may include: a processor 501, a network interface 504, and a memory 505. In addition, the above computer device 50 may further include: a user interface 503 and at least one communication bus 502. Among them, the communication bus 502 is used to realize the connection and communication between these components. Among them, the user interface 503 may include a display screen (Display) and a keyboard (Keyboard). Optionally, the user interface 503 may further include a standard wired interface and a wireless interface. The network interface 504 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 505 may be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. Optionally, the memory 505 may also be at least one storage device located far from the aforementioned processor 501. As Figure 5 shown, the memory 505, as a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a device control application program.

[0158] In Figure 5 the computer device 50 shown in the figure, the network interface 504 can provide network communication functions; while the user interface 503 is mainly used to provide an input interface for users; and the processor 501 can be used to call the device control application program stored in the memory 505 to achieve:

[0159] When an upload event for target video data is detected, obtain the target video data;

[0160] Obtain a target policy associated with the target video data, and predict whether the target video data belongs to a popular video based on the target policy. The target policy includes at least one of a video hit policy and a video prediction policy;

[0161] If the target video data belongs to a popular video, perform transcoding processing on the target video data to obtain transcoded video data;

[0162] When a request for obtaining the target video data is received, send the transcoded video data to the terminal device.

[0163] It should be understood that the computer device 50 described in the embodiments of the present application may execute the descriptions of the above data processing methods in the corresponding embodiments described above Figure 2 and Figure 3 and may also execute the descriptions of the above data processing devices in the corresponding embodiments described above Figure 4 and will not be elaborated herein. In addition, the descriptions of the beneficial effects of adopting the same method will not be elaborated either.

[0164] In the embodiment of the present application, target video data is obtained, and a target policy is obtained. Based on the target policy, it is predicted whether the target video data belongs to a popular video. If the target video data belongs to a popular video, the target video data is transcoded to obtain transcoded video data. When a request for obtaining the target video data is received, the transcoded video data is sent to the terminal device. Since this method can automatically predict the obtained target video data, the prediction efficiency of popular videos can be improved, and thus the data processing efficiency can be enhanced. Further, the target policy includes at least one of a video hit policy and a video prediction policy. When the target policy includes the video hit policy, it is possible to quickly determine whether the target video data is a popular video based on the video hit policy, improving the prediction efficiency. When the target policy includes the video prediction policy, it is possible to predict whether the target video data is a popular video based on the video prediction policy, improving the data prediction accuracy. When the target policy includes both the video hit policy and the video prediction policy, it is possible to preliminarily determine whether the target video data is a popular video based on the video hit policy, improving the data prediction accuracy. If it is preliminarily determined that the target video data does not belong to a popular video, the video prediction policy can be further used to perform a secondary determination on the target video data to determine whether the target video data belongs to a popular video, further improving the video prediction accuracy. In addition, by predicting popular videos and transcoding the popular videos, the transcoded video data can be directly obtained subsequently, which can save bandwidth and improve the data transmission efficiency.

[0165] The embodiment of the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer is caused to execute the method in the foregoing embodiment. The computer may be a part of the computer device mentioned above. For example, it is the aforementioned processor 501. As an example, the program instructions may be deployed to be executed on a computer device, or be deployed to be executed on multiple computer devices located at one place. Or, the program instructions may be executed on multiple computer devices distributed at multiple places and interconnected through a communication network. The multiple computer devices distributed at multiple places and interconnected through a communication network may form a blockchain network.

[0166] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it may include the processes of the embodiments of the above methods. Among them, the storage medium may be a magnetic disk, an optical disc, a read-only memory (ROM), or a random access memory (RAM), etc.

[0167] The above disclosure is only for the preferred embodiments of the present application. Of course, it cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application still fall within the scope covered by the present application.

Claims

1. A data processing method, characterized in that, The method includes: Obtaining target video data; Obtaining a target policy, and predicting whether the target video data belongs to a popular video based on the target policy. The target policy includes a video hit policy and a video prediction policy. The video hit policy is used to predict whether the target video data belongs to a popular video according to the feature data of the user. The video prediction policy is used to predict whether the target video data belongs to a popular video according to a video prediction model. The video hit policy and the video prediction policy refer to the policies associated with the category of the target video data obtained from multiple policies. One category of video data is associated with one policy. The video hit policy includes user attention and / or a whitelist. The prediction based on the target policy means: based on whether the user attention of the target user corresponding to the target video data is greater than the attention threshold and / or whether the target user hits the whitelist. If the attention of the target user is less than or equal to the attention threshold and / or the target user does not hit the whitelist, then the target video data is predicted using a target video prediction model, and a video score corresponding to the video features of the target video data and a user score corresponding to the user features are obtained respectively; obtaining the video category features of the target video data, and using the target video prediction model to respectively determine a first weight corresponding to the video features and a second weight corresponding to the user features based on the video category features. The video category features are used to indicate the category to which the target video data belongs; using the target video prediction model to determine the video prediction score based on the video score, the first weight, the user score, and the second weight, and determining whether the target video data belongs to a popular video based on whether the video prediction score is higher than the score threshold. Different categories of video data have different attention thresholds, and different categories of video data have different whitelists. The target video prediction model is determined from multiple video prediction models based on the service demand data corresponding to the target video data. The service demand data includes at least one of the number of popular videos, the prediction latency of popular videos, and the playback coverage rate of popular videos. Each video prediction model focuses on different aspects. The video category features can be used to reflect the degree of attention to the features of all aspects of the target video data; If the target video data belongs to a popular video, then perform transcoding processing on the target video data to obtain transcoded video data; When a request for obtaining the target video data is received, send the transcoded video data to the terminal device.

2. The method according to claim 1, characterized in that, The target policy includes a video hit policy. The video hit policy includes user attention and / or a whitelist. The user attention is used to reflect the degree of attention of the user. The whitelist includes at least one user; The predicting whether the target video data belongs to a popular video based on the target policy includes: Determine whether the user attention of the target user corresponding to the target video data is greater than the attention threshold and / or whether the target user hits the whitelist; If the attention of the target user is greater than the attention threshold and / or the target user hits the whitelist, determine that the target video data belongs to a popular video; If the attention of the target user is less than or equal to the attention threshold and / or the target user does not hit the whitelist, determine that the target video data does not belong to a popular video.

3. The method according to claim 1 or 2, characterized in that, The target strategy includes a video prediction strategy, and the video prediction strategy includes multiple video prediction models, and the video prediction model is used to predict whether the target video data belongs to a popular video; The obtaining of the target strategy and predicting whether the target video data belongs to a popular video based on the target strategy includes: Obtain the service demand data corresponding to the target video data, and determine the target video prediction model from the multiple video prediction models based on the service demand data; Use the target video prediction model to predict the target video data to obtain a video prediction result, and the video prediction result includes a video prediction score; If the video prediction score is higher than the score threshold, determine that the target video data belongs to a popular video; If the video prediction score is lower than or equal to the score threshold, determine that the target video data does not belong to a popular video.

4. The method according to claim 3, wherein The video features include video play volume and video interaction volume, and the user features include user attention and user play volume.

5. The method according to claim 3, characterized in that, The method further includes: Obtain multiple sample video data and sample labels; Use the initial video prediction model to predict each sample video data in the multiple sample video data to obtain a sample prediction result for each sample video data; Based on the sample prediction result of each sample video data and the sample label, determine the model deviation feature for the initial video prediction model; Train the initial video prediction model based on the model deviation feature to obtain the target video prediction model.

6. The method according to claim 1, characterized in that Before predicting whether the target video data belongs to a popular video based on the target strategy, the method further includes: Compare the target video data with the reference video data in the video database to determine the comparison result for the target video data, and the comparison result is used to indicate whether the target video data is duplicate video data; If the comparison result indicates that the target video data does not match each reference video data in the video database, perform the step of predicting whether the target video data belongs to a popular video based on the target strategy; If the comparison result indicates that the target video data matches the target reference video data in the video database, determine whether the target video data belongs to a popular video based on the video label corresponding to the target reference video data.

7. The method according to any one of claims 4-6, characterized in that, After transcoding the target video data to obtain transcoded video data, the method further includes: Push the transcoded video data to at least one terminal device; Detect operation instructions of the at least one terminal device for the transcoded video data within a target time period, where the operation instructions are used for at least one of playing, liking, favoriting, and forwarding the transcoded video data; If the data operation volume corresponding to the operation instructions is less than a quantity threshold, adjust the target policy.

8. A data processing device, characterized in that, The device includes: A data acquisition unit, configured to acquire target video data; A data prediction unit, configured to acquire a target policy, and predict whether the target video data belongs to a popular video based on the target policy. The target policy includes a video hit policy and a video prediction policy. The video hit policy is used to predict whether the target video data belongs to a popular video according to user feature data, and the video prediction policy is used to predict whether the target video data belongs to a popular video according to a video prediction model. The video hit policy and the video prediction policy refer to policies associated with the category of the target video data obtained from multiple policies, and one category of video data is associated with one policy. The video hit policy includes user attention and / or a whitelist. The prediction based on the target policy means: based on whether the user attention of the target user corresponding to the target video data is greater than an attention threshold and / or whether the target user hits the whitelist. If the attention of the target user is less than or equal to the attention threshold and / or the target user does not hit the whitelist, then predict the target video data based on a target video prediction model, and respectively obtain a video score corresponding to the video features of the target video data and a user score corresponding to the user features; obtain the video category features of the target video data, and use the target video prediction model to respectively determine a first weight corresponding to the video features and a second weight corresponding to the user features based on the video category features. The video category features are used to indicate the category to which the target video data belongs; use the target video prediction model to determine the video prediction score based on the video score, the first weight, the user score, and the second weight, and determine whether the target video data belongs to a popular video based on whether the video prediction score is higher than a score threshold. Different categories of video data have different attention thresholds and different whitelist. The target video prediction model is determined from multiple video prediction models based on the service requirement data corresponding to the target video data. The service requirement data includes at least one of the number of popular videos, the prediction latency of popular videos, and the playback coverage rate of popular videos. Each video prediction model focuses on different aspects, and the video category features can be used to reflect the degree of attention to various aspects of the target video data; A data transcoding unit, configured to perform transcoding processing on the target video data to obtain transcoded video data if the target video data belongs to a popular video; A data sending unit, configured to send the transcoded video data to a terminal device when receiving an acquisition request for the target video data.

9. A computer device, characterized in that, Includes: A processor, a memory, and a network interface; The processor is connected to the memory and the network interface. Among them, the network interface is used to provide data communication functions, the memory is used to store program codes, and the processor is used to call the program codes so that the computer device executes the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which is adapted to be loaded and executed by a processor so that a computer device having the processor executes the method according to any one of claims 1-7.

11. A computer program product, characterized in that, The computer program product includes computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1-7 is implemented.

Images