CN113569150B

CN113569150B - Media content preloading method, model building method and related equipment

Info

Publication number: CN113569150B
Application number: CN202110915696.0A
Authority: CN
Inventors: 张行功; 许智敏; 张颢丹; 郭宗明; 马茜; 王悦
Original assignee: Peking University; Beijing ByteDance Network Technology Co Ltd
Current assignee: Peking University; Beijing ByteDance Network Technology Co Ltd
Priority date: 2021-08-10
Filing date: 2021-08-10
Publication date: 2024-04-19
Anticipated expiration: 2041-08-10
Also published as: CN113569150A

Abstract

The present disclosure relates to a media content preloading method, a model building method and related devices, the media content preloading method comprising: acquiring historical viewing information of a user, play state information of electronic equipment and attribute information of media content to be recommended; inputting historical viewing information, play state information and attribute information of a user into a recommendation model, obtaining recommendation result information output by the recommendation model, determining user experience indexes to be optimized by the recommendation model based on the historical viewing information of the user, and outputting recommendation result information for optimizing the user experience indexes according to the play state information and the attribute information; and determining target media content from the media content to be recommended according to the recommendation result information to perform preloading. The method and the device can effectively improve the user experience quality of the user when watching the media content.

Description

Media content preloading method, model building method and related equipment

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to a media content preloading method, a model building method, and related devices.

Background

With the popularity of mobile networks and the development of streaming media technologies, mobile media application products (e.g., tremble, tikTok, fast-hander, etc.) have become a major tool for entertainment in life, wherein the media content may be audio, video, a combination thereof, and so on.

In the process of watching the media content, the user may switch from the currently played media content to the next media content at any time, so that the mobile terminal needs to preload the next media content to be played, and the next media content can be ensured to be played quickly. But there is currently no better method of preloading media content, resulting in a poor user experience.

Disclosure of Invention

This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.

In a first aspect, the present disclosure provides a media content preloading method, comprising:

Acquiring historical viewing information of the user, playing state information of the electronic equipment and attribute information of media content to be recommended;

Inputting the historical viewing information, the playing state information and the attribute information of the user into a recommendation model, obtaining recommendation result information output by the recommendation model, determining user experience indexes to be optimized by the recommendation model based on the historical viewing information of the user, and outputting recommendation result information for optimizing the user experience indexes according to the playing state information and the attribute information;

And determining target media content from the media content to be recommended according to the recommendation result information to perform preloading.

In a second aspect, the present disclosure provides a model building method, the model building method comprising:

acquiring a historical viewing information sample of a user, a playing state information sample of electronic equipment and an attribute information sample of media content to be recommended;

determining user experience indexes of the user according to the play state information samples;

And training a recommendation model by taking the user experience index as the tag information of the historical viewing information sample, the play state information sample and the attribute information sample, so that the recommendation model determines the user experience index to be optimized based on the historical viewing information sample, and outputting recommendation result information for optimizing the user experience index according to the play state information sample and the attribute information sample, wherein the recommendation result information is used for determining target media content to be preloaded from the media content to be recommended.

In a third aspect, the present disclosure provides a media content preloading apparatus, comprising: the system comprises an information acquisition module, a recommendation result information acquisition module and a preloading module.

Wherein: the information acquisition module is used for acquiring the historical viewing information of the user, the playing state information of the electronic equipment and the attribute information of the media content to be recommended;

the recommendation result information acquisition module is used for inputting the historical viewing information, the playing state information and the attribute information of the user into a recommendation model, acquiring recommendation result information output by the recommendation model, determining a user experience index to be optimized by the recommendation model based on the historical viewing information of the user, and outputting recommendation result information for optimizing the user experience index according to the playing state information and the attribute information;

And the preloading module is used for determining target media content from the media content to be recommended according to the recommendation result information to preload.

In a fourth aspect, the present disclosure provides a model building apparatus, comprising: the system comprises a sample acquisition module, a user experience index determination module and a training module.

Wherein: the sample acquisition module is used for acquiring a historical viewing information sample of a user, a play state information sample of the electronic equipment and an attribute information sample of media content to be recommended;

the user experience index determining module is used for determining the user experience index of the user according to the play state information sample;

The training module is used for taking the user experience index as the tag information of the historical viewing information sample, the play state information sample and the attribute information sample, training a recommendation model, enabling the recommendation model to determine the user experience index to be optimized based on the historical viewing information sample, outputting recommendation result information for optimizing the user experience index according to the play state information sample and the attribute information sample, and determining target media content to be preloaded from the media content to be recommended.

In a fifth aspect, the present disclosure provides a computer readable medium having stored thereon a computer program, characterized in that the program when executed by a processing device implements the steps of the method according to the first aspect.

In a sixth aspect, the present disclosure provides a computer readable medium having stored thereon a computer program, characterized in that the program when executed by a processing device implements the steps of the method according to the second aspect.

In a seventh aspect, the present disclosure provides an electronic device comprising:

a storage device having a computer program stored thereon;

Processing means for executing said computer program in said storage means to carry out the steps of the method of the first aspect.

In an eighth aspect, the present disclosure provides an electronic device comprising:

a storage device having a computer program stored thereon;

Processing means for executing said computer program in said storage means to carry out the steps of the method of the second aspect.

According to the technical scheme, in the process of playing the media content by the electronic equipment, the historical viewing information of the user, the playing state information of the electronic equipment and the attribute information of the media content to be recommended can be obtained, the historical viewing information of the user, the playing state information and the attribute information are input into the recommendation model, the recommendation result information output by the recommendation model is obtained, wherein the recommendation model determines the user experience index to be optimized based on the historical viewing information of the user, and the recommendation result information for optimizing the user experience index is output according to the playing state information and the attribute information.

Additional features and advantages of the present disclosure will be set forth in the detailed description which follows.

Drawings

The above and other features, advantages, and aspects of embodiments of the present disclosure will become more apparent by reference to the following detailed description when taken in conjunction with the accompanying drawings. The same or similar reference numbers will be used throughout the drawings to refer to the same or like elements. It should be understood that the figures are schematic and that elements and components are not necessarily drawn to scale. In the drawings:

Fig. 1 is a schematic view of an application environment of a media content preloading method and a model building method according to an embodiment of the present disclosure.

Fig. 2 is a flow chart of a model building method according to an embodiment of the present disclosure.

Fig. 3 is a flow chart of a model building method according to another embodiment of the present disclosure.

Fig. 4 is a flow chart illustrating a media content loading method according to an embodiment of the present disclosure.

Fig. 5 is a flowchart of a media content loading method according to another embodiment of the present disclosure.

Fig. 6 is a network structure diagram of a recommendation model provided in an embodiment of the present disclosure.

Fig. 7 is a schematic flowchart of an implementation of a media content preloading method according to an embodiment of the present disclosure in practical applications.

Fig. 8 is a schematic structural diagram of a media content preloading device according to an embodiment of the present disclosure.

Fig. 9 is a schematic structural view of a model building apparatus provided in an embodiment of the present disclosure.

Fig. 10 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure.

Detailed Description

Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While certain embodiments of the present disclosure have been shown in the accompanying drawings, it is to be understood that the present disclosure may be embodied in various forms and should not be construed as limited to the embodiments set forth herein, but are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustration purposes only and are not intended to limit the scope of the present disclosure.

It should be understood that the various steps recited in the method embodiments of the present disclosure may be performed in a different order and/or performed in parallel. Furthermore, method embodiments may include additional steps and/or omit performing the illustrated steps. The scope of the present disclosure is not limited in this respect.

The term "including" and variations thereof as used herein are intended to be open-ended, i.e., including, but not limited to. The term "based on" is based at least in part on. The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments. Related definitions of other terms will be given in the description below.

It should be noted that the terms "first," "second," and the like in this disclosure are merely used to distinguish between different devices, modules, or units and are not used to define an order or interdependence of functions performed by the devices, modules, or units.

It should be noted that references to "one", "a plurality" and "a plurality" in this disclosure are intended to be illustrative rather than limiting, and those of ordinary skill in the art will appreciate that "one or more" is intended to be understood as "one or more" unless the context clearly indicates otherwise.

The names of messages or information interacted between the various devices in the embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

With the popularization of mobile networks and the development of streaming media technology, mobile media application products become a main tool for entertainment in life, wherein the media content can be audio, video, a combination thereof, and the like. In multimedia applications, one typical scenario is a feed media stream, i.e. a user can switch to the next media content by sliding down or to the last media content by sliding up, where the media content may be played and displayed automatically, and another typical scenario is a mobile phone window displaying a plurality of media contents, where the user can browse different media contents by sliding up and down and click to view the different media contents, and in these scenarios, the server side will forward a list of recommended media contents (e.g. containing 6 or more media contents) to the client side in advance.

In the process of watching media content, a user may switch to the next media content at any time, if the media content after currently watching the media content in the recommended media content list is not preloaded, the user has a relatively long initial frame time when watching the next media content, and meanwhile, extra clamping is caused under the condition of insufficient network bandwidth, wherein the initial frame time refers to the time difference from the starting time of clicking to play the media content or automatically playing the media content to the first frame picture of the media content after switching to the fixed media content.

In order to reduce the first frame time and the blocking, the current common method is to perform a preloading method with fixed size/duration, and preload the subsequent media content in the recommendation list with fixed byte number or fixed duration according to the sequence after downloading the media content being watched by the user, wherein the subsequent media content can be downloaded in parallel; in addition, many products do not employ a pre-load strategy.

In some examples, existing methods target optimization of the metrics of the stuck, the first frame time, the fluency, etc. In actual optimization, linear combination of a plurality of indexes is taken as an optimization target. However, there may be a conflict between optimizations of different metrics, e.g., preferential preloading of media content may reduce the first frame time, but increase the churning.

Therefore, in the method, the optimization target cannot be guaranteed to be consistent with the quality of experience of the user by simply and linearly combining various indexes, so that the media content recommendation result obtained by simply and linearly combining various indexes cannot be guaranteed to be good in quality of experience of the user.

In other examples, the machine learning-based preloading method has the problem of generalizing performance, namely, the effect of the algorithm model is strongly related to the distribution of the training data set, and is difficult to use on other distributed data sets. Reinforcement learning is a commonly used machine learning algorithm that can be trained on a network bandwidth dataset and a user viewing behavior dataset to derive a preloading strategy. The "fast-sliding" behavior (the user quickly skips over a part of uninteresting media content to find the media content of interest) and the "slow-sliding" behavior (the user finishes watching each recommended media content) in the user watching process are two behaviors with larger distribution difference.

In one aspect, a model trained based on a fast-sliding dataset may tend to preload multiple subsequent media content, each for a shorter duration, maintaining a shorter buffer for the media content being viewed. This fast-sliding model can cause more stuck when applied to slow-sliding datasets.

On the other hand, models trained based on slow-slide data sets may tend to maintain a longer buffer for media content being viewed, preloading a smaller number of subsequent media content, preloading each media content for a longer period of time. This slow-slip model results in a large initial frame time when applied to a fast-slip data set.

On the other hand, if the model is trained by blending two data sets together and the number of data from one of the blended data sets is dominant, the algorithm model will tend to adapt to such data set, resulting in the same effect as previously described. If the data amounts of the two data sets are comparable, the model characteristics of training will be intermediate between the two models, i.e., the buffer size maintained is intermediate between the fast and slow sliding models, and the preloaded media content amounts and preloaded sizes are intermediate between the fast and slow sliding models. Such a model is not as fast-sliding on a fast-sliding dataset and not as slow-sliding on a slow-sliding dataset.

It can be seen that the preloading of media content by the fixed model is not flexible to implement and the user experience is poor.

In view of the above problems, the present disclosure provides a media content preloading method, a model building method and related devices, which can effectively improve the experience quality of a user when watching media content.

In the following, referring to fig. 1, the media content preloading method and the model building method provided by the present disclosure may be applied to a media content recommendation system shown in fig. 1, where the media content recommendation system may include an electronic device 110 and a server 120 of a user, where the electronic device 110 may be configured with a processor, a media content playing device, a communication device, and so on, and the electronic device 110 may establish a communication link with the server 120 through the communication device. The server 120 may perform functions such as data receiving, data processing, data transmitting, and data storing, and specifically, the server 120 may receive data generated by the electronic device 110, process the data generated by the electronic device 110, and then send the processed data to the electronic device 110. Alternatively, the media content preloading method and the model building method may be separately applied to the electronic device 110 in the media content recommendation system, may be separately applied to the server 120 in the media content recommendation system, and may also be simultaneously applied to the server 120 and the electronic device 110 in the media content recommendation system.

Referring to fig. 2, which illustrates a model construction method provided by an embodiment of the present disclosure, the model construction method may be applied to the server as illustrated in fig. 1, a model constructed by the model construction method may be used to recommend media content, and the model construction method may include the steps of:

101. and acquiring a historical viewing information sample of the user, a playing state information sample of the electronic equipment and an attribute information sample of the media content to be recommended.

In some embodiments, the server (hereinafter may be referred to as a server) establishes a communication connection with an electronic device (hereinafter may be referred to as a client) used by the user, receives and records viewing information generated when the user views media content on the client, and playing state information of the client in real time, and then may use the viewing information recorded in a period of time as a historical viewing information sample and the recorded playing state information as a playing state information sample. The database of the server can be pre-stored with a plurality of attribute information samples of the media content to be recommended, and the attribute information samples can be directly called from the database when required.

Wherein, the historical viewing information can comprise playing state information of the historical media content watched by the user and/or historical behavior information of the historical media content watched by the user.

Optionally, the historical behavior information includes at least one of a viewing duration of the user, a duration of the viewed media content, an exit operation when the viewed media content appears to be a clip, and an exit operation in waiting for a first frame time of the media content.

Optionally, the attribute information of the media content to be recommended includes at least one of a duration of the media content to be recommended and a code rate of the media content to be recommended.

Optionally, the playing state information includes: at least one of a duration of the remaining media content in the corresponding buffer of the currently played media content, a duration of the remaining media content in the corresponding buffer of the unreported media content, a number and duration of the jams that occur in the playing process of the currently played media content, a first frame time of the currently played media content, download information of the currently played media content, and download information of the unreported media content. Wherein the download information of the media content characterizes whether the download of the media content is completed.

102. And determining user experience indexes of the user according to the play state information samples.

The user experience index may be used to characterize the user experience quality of the user or the satisfaction degree of the user in the playing state corresponding to the playing state information sample. For example, the user experience index may be a numerical value, which may be positively correlated with the quality of experience of the user. For another example, the user experience index may also be data made by the user that may reflect a series of behavior actions experienced by the user, such as a rate of exit of the user from playing media content, where a higher or higher rate of exit may indicate that the user experience is poor. Such as the viewing duration of a user viewing media content, when the light viewing duration is shorter or shorter, the user experience may be indicated as being poor. Thus, the user experience metrics may include metrics such as drop-out rate, viewing duration, etc.

In some embodiments, the server may fit a user experience quality function in advance according to a large number of historical viewing information samples, and since the historical viewing information includes playing state information of historical media content watched by the user and historical behavior information of the historical media content watched by the user, the user experience quality function obtained by the fitting may reflect a relationship between the playing state information of the media content and the user viewing behavior information, and the user viewing behavior may directly reflect a user experience index, so that the user experience quality function may reflect a relationship between the playing state information of the media content and the user experience index. After the establishment of the user experience quality function is completed, the server can substitute the play state information sample into the user experience quality function, and the user experience index is obtained through calculation. For example, the user experience index may be a playback rate or/and a duration in a playing state, such as an increase in the number of times of a pause, an increase in the length of a pause, an increase in the first frame time, or a change in the code rate of the media content.

103. And training a recommendation model by taking the user experience index as tag information of the historical viewing information sample, the playing state information sample and the attribute information sample, so that the recommendation model determines the user experience index to be optimized based on the historical viewing information sample, and outputting recommendation result information for optimizing the user experience index according to the playing state information sample and the attribute information sample, wherein the recommendation result information is used for determining target media content to be preloaded from media content to be recommended.

For example, the server may determine that the user experience index does not meet the preset condition as the user experience index to be optimized, for example, if the exit rate when the number of the jams increases is greater than the exit rate threshold, the exit rate when the number of the jams increases may be determined as the user experience index to be optimized, and it may be determined that the cause that the user experience index to be optimized does not reach the standard is the increase of the number of the jams according to the user experience index to be optimized. The reasons for the user experience index to be optimized not reaching the standard may be one or more.

The trained recommendation model can output recommendation result information capable of optimizing the user experience index according to the user experience index to be optimized, the play state information sample and the attribute information sample, and preloaded target media content is obtained, for example, the reason that the user experience index to be optimized does not reach the standard is that the initial frame time is long, so that the preloaded target media content determined according to the recommendation result information output by the trained recommendation model can overcome the problem that the initial frame time is long in play, and therefore user experience quality of a user is effectively improved.

It can be seen that, in this embodiment, by acquiring a historical viewing information sample of a user, a play status information sample of an electronic device, and an attribute information sample of media content to be recommended. Determining user experience indexes of a user according to a play state information sample, taking the user experience indexes as tag information of a historical viewing information sample, a play state information sample and an attribute information sample, training a recommendation model, enabling the recommendation model to determine user experience indexes to be optimized based on the historical viewing information sample, outputting recommendation result information capable of optimizing the user experience indexes according to the play state information sample and the attribute information sample, and determining target media content to be preloaded from media content to be recommended, wherein the recommendation result information is used for determining target media content capable of overcoming the problem of the root cause of the user experience according to the root cause of the user experience, and determining the target media content capable of overcoming the problem of the root cause of the user experience according to the play state information sample and the attribute information sample of the media content to be recommended.

Referring to fig. 3, a model building method according to another embodiment of the present disclosure is shown, which may be applied to the server shown in fig. 1, and may include the steps of:

201. And initializing a simulation environment based on the preset network bandwidth data and the real viewing behavior information of the user.

202. And simulating a process of watching the media content by the user through the electronic equipment in a simulation environment, and acquiring a historical watching information sample of the user, a playing state information sample of the electronic equipment and an attribute information sample of the media content to be recommended, which are generated in the process.

203. And determining user experience indexes of the user according to the play state information samples.

204. And training a recommendation model by taking the user experience index as tag information of the historical viewing information sample, the playing state information sample and the attribute information sample, so that the recommendation model determines the user experience index to be optimized based on the historical viewing information sample, and outputting recommendation result information for optimizing the user experience index according to the playing state information sample and the attribute information sample, wherein the recommendation result information is used for determining target media content to be preloaded from media content to be recommended.

In some embodiments, the recommendation model includes an upper model for learning a user experience index to be optimized based on the historical viewing information samples, and a plurality of lower models for learning an influence relationship of the play state information samples and the attribute information samples on one user experience index.

For example, the trained upper layer model may output a user experience index to be optimized according to the input historical viewing information, a target lower layer model for specifically solving the reason that the user experience index to be optimized does not reach the standard may be determined from a plurality of lower layer models according to the user experience index to be optimized, for example, the reason that the user experience index to be optimized does not reach the standard is that the number of the jams is increased, then the target lower layer model may be a model specially used for processing the increase of the number of the jams, after the target lower layer model is determined, the trained target lower layer model may output recommendation result information according to the input playing state information and the media content attribute information to be recommended, and the target media content determined according to the recommendation result information may effectively reduce the number of the jams.

In some embodiments, the network parameters of the recommendation model include a number of activations set by the upper layer model for each lower layer model, the number of activations being used to characterize a number of repetitions of media content recommendation by the recommendation model each time the lower layer model is employed, the method further comprising:

205. And after each training, re-acquiring a new play state information sample, and determining a new user experience index according to the new play state information sample.

206. And updating the network parameters of the recommendation model according to the user experience index and the change condition of the new user experience index.

In the following, steps 201 to 205 will be described as a whole, and by way of example, in practice, the server may implement training of the recommendation model by performing the following steps:

the first step, initializing an upper layer model parameter and a plurality of lower layer model parameters.

And secondly, selecting a piece of network bandwidth data and a piece of user watching behavior data to initialize the simulation environment.

And thirdly, selecting a lower model to be activated and the number S of activated steps by the upper model according to the input.

And step four, selecting media content to be downloaded by the lower layer model according to the input, repeating the step S, if the user does not finish watching, executing the step three, and if the user finishes watching, executing the step five.

And fifthly, updating network parameters of the model by using a reinforcement learning algorithm according to input and output in the simulation process and changes of user experience indexes. And returns to the fourth step.

In this embodiment, the simulation environment is initialized by being based on preset network bandwidth data and real viewing behavior information of the user. And simulating a process of watching the media content by the user through the electronic equipment in a simulation environment, and acquiring a historical watching information sample of the user, a playing state information sample of the electronic equipment and an attribute information sample of the media content to be recommended, which are generated in the process. Therefore, sample data can be acquired in a simulation environment, the tedious process caused by acquiring real data as the sample data is avoided, and the model training efficiency is improved.

Referring to fig. 4, a media content loading method according to an embodiment of the present disclosure is shown, which may be applied to the server or the electronic device shown in fig. 1, and may include the following steps:

301. And acquiring historical viewing information of the user, playing state information of the electronic equipment and attribute information of media content to be recommended.

The media content loading method may be performed in a scenario where media content is played electronically and next media content is played automatically, for example, in a scenario where a user views a jittery audio and short video, or in a scenario where a user plays a plurality of songs continuously through music playing software, for example.

In some examples, a server in communication with the electronic device may receive, in real-time, viewing information generated when a user views media content on the electronic device, playback state information of the client, and take the viewing information generated by the user over a specified period of time as historical viewing information during the user views the media content through the electronic device. The media content to be recommended and the attribute information of the media content to be recommended may be stored in a database of the server in advance, and may be called from the database when in use, or the server may send request information to the media content application platform to instruct the media content application platform to feed back the media content to be recommended and the attribute information of the media content to be recommended.

Wherein the historical viewing information comprises playing state information of the historical media content watched by the user and/or historical behavior information of the historical media content watched by the user.

Wherein the historical behavior information includes at least one of a viewing time period of the user, a time period of the viewed media content, an exit operation when the viewed media content appears to be a clip, and an exit operation in waiting for a first frame time of the media content.

The attribute information of the media content to be recommended comprises at least one of duration of the media content and code rate of the media content.

Wherein, the play status information includes: at least one of a duration of the remaining media content in the corresponding buffer of the currently played media content, a duration of the remaining media content in the corresponding buffer of the unreported media content, a number and duration of the jams that occur in the playing process of the currently played media content, a first frame time of the currently played media content, download information of the currently played media content, and download information of the unreported media content. Wherein the download information of the media content characterizes whether the download of the media content is completed.

302. And inputting the historical viewing information, the playing state information and the attribute information of the user into a recommendation model, and obtaining recommendation result information output by the recommendation model, wherein the recommendation model determines user experience indexes to be optimized based on the historical viewing information of the user, and outputs recommendation result information for optimizing the user experience indexes according to the playing state information and the attribute information.

The user experience index to be optimized is determined based on play state information and/or historical behavior information of the historical media content, and the play state information is used for representing excessive click times or excessively long click time, so that the withdrawal rate is determined to be the user experience index to be optimized, the historical behavior information is used for representing excessively low watching time, and the user experience index to be optimized can be time.

For example, if the user experience index to be optimized is a duration, recommendation result information capable of optimizing the duration may be output according to the play state information and the attribute information, for example, when the user experience index to be optimized is a duration, it indicates that the duration of watching the media content by the user is shorter, and then the recommendation result information may be used to enable the user to increase the duration of watching the media content. Optionally, the recommendation result information may be target attribute information of the media content to be recommended, and when the media content to be recommended corresponding to the target attribute information is preloaded and played in the current playing state of the electronic device, the viewing duration of the user may be optimized.

303. And determining target media content from the media content to be recommended according to the recommendation result information to perform preloading.

In some embodiments, a list of media contents to be recommended may be stored in advance in the server, the list of media contents to be recommended including a plurality of media contents to be recommended and attribute information of each media content to be recommended. In the above example, if the recommendation result information may be target attribute information of the media content to be recommended, the server may determine, as the target media content, the media content in the media content list to be recommended that matches the target attribute information, and send the target media content to the client, so as to instruct the client to preload the target media content.

In this embodiment, in a process that a user views media content through electronic equipment, historical viewing information of the user, playing state information of the electronic equipment and attribute information of the media content to be recommended are obtained, the historical viewing information of the user, the playing state information and the attribute information are input into a recommendation model, recommendation result information output by the recommendation model is obtained, wherein the recommendation model can determine a user experience index to be optimized based on the historical viewing information of the user, and output recommendation result information capable of optimizing the user experience index according to the playing state information and the attribute information.

Referring to fig. 5, a media content preloading method according to another embodiment of the present disclosure is shown, which may be applied to the server or the electronic device shown in fig. 1, and may include the following steps:

401. And acquiring historical viewing information of the user, playing state information of the electronic equipment and attribute information of media content to be recommended.

The specific embodiment of step 401 may refer to step 301, and thus will not be described herein.

In some embodiments, the recommendation model includes an upper model and a plurality of lower models, where the plurality of lower models are in one-to-one correspondence with the plurality of user experience indexes, and each lower model learns in advance an influence relationship between play state information and attribute information of the media content sample on the user experience indexes corresponding to the lower models.

For example, the network structure of the recommendation model may be as shown in fig. 6, where the network structure of the recommendation model is composed of two layers of networks, and may specifically include: an upper model neural network and K lower model neural networks.

The upper layer model is used for selecting the most suitable lower layer model according to the historical viewing information of the user.

Different ones of the underlying networks are different preload strategies, each preload strategy selecting media content to be downloaded based on current state (buffer size, duration, code rate, whether or not full download of media content is being played and not being played, number of clicks and time that current media content has occurred, first frame time of current media content). The size of the download may be set to a fixed value or may be output from the network.

For example, the plurality of lower models may include: the method comprises the steps of a lower model corresponding to a user experience index with a user experience index of a withdrawal rate, a lower model corresponding to a user experience index with a user experience index of a viewing duration, a lower model corresponding to a user experience index with a user experience index of a short switching duration (such as a fast sliding model), and a lower model corresponding to a user experience index with a user experience index of a long switching duration (such as a slow sliding model), wherein the switching duration refers to the duration of a user switching from one video to the next video.

402. And inputting the historical viewing information into an upper model, and determining a target lower model from a plurality of lower models according to the output result of the upper model.

By adopting the above example, after the history viewing information is input into the upper model, the upper model may output the user experience index to be optimized, and according to the user experience index to be optimized, a target lower model corresponding to the user experience index to be optimized may be determined from the plurality of lower models.

403. And inputting the play state information and the attribute information into the target lower layer model to obtain recommendation result information output by the target lower layer model.

In some embodiments, the server may input the play status information and the attribute information of the media content to be recommended into the target lower layer model, and the target lower layer model may output recommendation result information by combining the play status information and the attribute information of the media content to be recommended with respect to the user experience index to be optimized. The recommendation result information may include a code rate, a size, a name, a resolution, a duration, etc. of the target media content.

404. And determining target media content from the media content to be recommended according to the recommendation result information to perform preloading.

In some embodiments, the output result of the upper layer model includes a number of activations corresponding to the target lower layer model, where the number of activations is a number of times determined in the training process for enhancing a user experience index to be optimized to an expected index value, and the preloading method of the media content further includes:

405. After the target media content is determined to be preloaded from the media content to be recommended according to the recommendation result information, the following operations are repeatedly executed aiming at the target lower layer model until the execution times reach the activation times.

406. And re-acquiring the playing state information of the electronic equipment and the attribute information of the media content to be recommended.

407. And inputting the new play state information and the new attribute information into the target lower model to obtain new recommendation result information.

408. And determining target media content from the media content to be recommended according to the new recommendation result information to perform preloading.

Optionally, before each re-executing step 406 to step 408, it further includes detecting whether the user behavior of the user meets a preset condition, for example, if the user exit client is not detected during the execution of step 406 to step 408, it may be determined that the user behavior meets the preset condition, and then the next process from step 406 to step 408 may be re-executed.

In this embodiment, after the target media content is determined from the media content to be recommended according to the recommendation result information and preloaded, the steps 406 to 408 are repeatedly executed for the target lower layer model until the execution times reach the activation times, and because the target activation times are times which are determined in the training process and can raise the user experience index to be optimized to the expected index value, it can be ensured that the user experience index can be effectively raised to the expected index value after repeated execution for many times, and when the user experience index can be effectively raised to the expected index value, it indicates that the user has higher experience quality.

In some implementations, the media content preloading method further includes:

When the number of executions reaches the number of activations, the historical viewing information of the user is acquired again, and the step of inputting the historical viewing information into the upper model is performed according to the new historical viewing information, that is, the step is returned to the step 402, and the steps 402 to 408 are performed again.

For example, when the number of activation times is S, after step 402 to step 408 have been repeated S times, that is, when S target media contents are determined to be preloaded through the currently determined target lower layer model, the user experience index to be optimized may be redetermined through the upper layer model, and then a new target lower layer model is redelected, so as to ensure that the user experience index to be optimized can be updated in real time, and continuously ensure that the user has higher user experience quality.

In some embodiments, the target lower layer model is a plurality, and the specific embodiment of step 403 may include: and inputting the play state information and the attribute information into each target lower layer model to obtain recommendation result information output by each target lower layer model.

Accordingly, a specific embodiment of step 404 may include: and determining target media content recommended by each recommendation result information from media content to be recommended to load.

For example, if the target lower layer model includes a lower layer model (hereinafter may be referred to as a first lower layer model) corresponding to a duration and a lower layer model (hereinafter may be referred to as a second lower layer model) corresponding to an exit rate, the server may input playing state information and attribute information of media content to be recommended into the first lower layer model and the second lower layer model, respectively, obtain first recommendation result information output by the first lower layer model and second recommendation result information output by the second lower layer model, determine the first target media content according to the first recommendation result information, determine the second target media content according to the second recommendation result information, and finally preload the first target media content and the second target media content. Optionally, priorities of the preset first lower layer model and the second lower layer model may be obtained, and whether the first target media content or the second target media content is loaded first or not may be determined according to the priorities.

The method for preloading the media content will be described below in conjunction with steps 401 to 408, and as shown in fig. 7, by way of example, in practical application, the execution flow of the method for preloading the media content may be as follows:

Step 1, a user opens a client.

And 2, loading a pre-trained recommendation model.

And step 3, obtaining upper model input.

By way of example, the upper model input may include historical viewing media content attributes (media content duration, user viewing duration, media content rate); the currently played media content corresponds to the duration of the remaining media content in the buffer; the unplayed media content corresponds to the duration of the remaining media content in the buffer.

And step 4, acquiring an upper model to activate a lower model and selecting the number of activated steps (target activation times).

And step 5, obtaining the input of the lower model.

Illustratively, the lower model input may include attributes of the media content in the recommended media content list (media content duration, media content rate); the currently played media content corresponds to the duration of the remaining media content in the buffer; the media content that is not played corresponds to the duration of the remaining media content in the buffer; whether media content being played and not being played is completely downloaded; the number and duration of the blocking in the current media content playing process; the first frame time of the current media content, etc.

In step6, the lower model selects the media content to be downloaded.

Step 7, downloading the media content.

And 8, judging whether the user exits the client.

And step 9, if the user exits the client, closing the client.

And step 10, if the user does not exit the client, judging whether the current activation step number reaches a set value, namely whether the target activation times are reached.

And 11, returning to the step 3 if the number of the current activation steps reaches the set value, and re-executing the input of the upper model.

And step 12, returning to the step 5 if the number of the current activation steps does not reach the set value, and re-acquiring the input of the lower model.

Alternatively, the recommendation model may be configured on the client, and the client recommendation model reasoning (online decision) starts running after the user opens the client. The client loads model parameters, acquires input required by an upper model, selects a lower model to be activated according to the input, and selects the number of steps to be activated. And then the client acquires the input required by the lower layer model, the activated lower layer model selects the media content to be downloaded according to the input, and the media content is delivered to a client download data module for downloading, wherein the downloading size can be set as a fixed value or can be used as the output of the lower layer model. When the client downloading module is interrupted (caused by reasons such as video switching by a user, client closing by the user, downloading ending and the like), if the user closes the client, the operation is decided to be stopped; if the activation steps of the lower layer model reach the set value, the input of the upper layer is required to be re-acquired and re-decided by the upper layer model; if the number of activation steps does not reach the set value, the lower model continues to make decisions.

The following is an experiment performed on a jittering user viewing data set and a bandwidth data set by using 3000 user viewing records, wherein each user viewing record comprises the viewing duration of all the viewing media contents from a user opening client to a user closing client, the media content duration, the media content code rate and the downloading network speed of each media content.

The method of preloading for comparison may include two methods:

1) The fixed pre-load, download the media content that the user is watching first, after downloading completely, download 800KB to other media content in the tabulation.

2) Preloading based on reinforcement learning.

The specific experimental results are shown in table 1:

TABLE 1

Compared with a fixed preloading scheme, the scheme has the advantages that the clamping proportion is reduced by 37%, the average clamping times are reduced by 38%, the first frame is reduced by 70%, and the user experience quality is improved by 4.7%. Compared with reinforcement learning preloading, the method has the advantages that the clamping and stopping proportion is reduced by 19%, the average clamping and stopping times are reduced by 34%, the first frame is reduced by 55%, and the user experience quality is improved by 4.4%.

In this embodiment, the historical viewing information is input into the upper model, the target lower model is determined from a plurality of lower models according to the output result of the upper model, the playing state information and the attribute information are input into the target lower model, and the recommendation result information output by the target lower model is obtained, so that the recommendation model is subjected to layered training, a proper lower model can be selected according to the actual playing state and the user viewing information, the use flexibility of the model is improved, and the situation that the user experience is poor due to the fact that the user experience is caused by the fact that the model is only used with the quick sliding model, the slow sliding model is only used, or the target recommendation content output by the model is mixed by the two models is avoided, so that the experience quality of the user is effectively improved.

Referring now to fig. 8, which illustrates a media content preloading device provided by an embodiment of the present disclosure, the media content preloading device 500 includes:

the information obtaining module 510 is configured to obtain historical viewing information of a user, playing state information of an electronic device, and attribute information of media content to be recommended.

The recommendation result information obtaining module 520 is configured to input historical viewing information, playing state information, and attribute information of a user into a recommendation model, obtain recommendation result information output by the recommendation model, determine a user experience index to be optimized based on the historical viewing information of the user, and output recommendation result information for optimizing the user experience index according to the playing state information and the attribute information.

The preloading module 530 is configured to determine, according to the recommendation result information, a target media content from the media contents to be recommended for preloading.

The recommendation result information obtaining module 520 includes:

The upper layer model activation submodule is used for inputting historical viewing information into the upper layer model and determining a target lower layer model from a plurality of lower layer models according to the output result of the upper layer model.

And the lower model activation submodule is used for inputting the play state information and the attribute information into the target lower model to obtain recommended result information output by the target lower model.

In some embodiments, the output of the upper layer model includes a number of activations of the corresponding target lower layer model, and the media content preloading device 500 further includes:

the repeated execution module is used for repeatedly executing the following operations aiming at the target lower layer model after determining the target media content to be recommended from the media content to be recommended to be preloaded according to the recommendation result information until the execution times reach the activation times:

and re-acquiring the playing state information of the electronic equipment and the attribute information of the media content to be recommended.

And inputting the new play state information and the new attribute information into the target lower model to obtain new recommendation result information.

And determining target media content from the media content to be recommended according to the new recommendation result information to perform preloading.

In some implementations, the media content preloading device 500 further includes:

And the lower model updating model is used for acquiring the historical viewing information of the user again when the execution times reach the activation times, and executing the step of inputting the historical viewing information into the upper model according to the new historical viewing information.

In some embodiments, the target lower layer model is a plurality of, and the recommendation result information obtaining module 520 is further configured to: and inputting the play state information and the attribute information into each target lower layer model to obtain recommendation result information output by each target lower layer model.

Correspondingly, the preloading module 530 is further configured to determine, from the media contents to be recommended, that the target media contents recommended by each recommendation result information are loaded.

In some embodiments, the historical viewing information includes play state information of historical media content viewed by the user and/or historical behavior information of the historical media content viewed by the user, and the user experience metrics to be optimized are determined based on the play state information and/or the historical behavior information of the historical media content.

In some implementations, the historical behavior information includes at least one of a viewing duration of the user, a duration of the viewed media content, an exit operation when the viewed media content appears to be a pause, an exit operation in waiting for a first frame time of the media content.

In some implementations, the attribute information of the media content to be recommended includes at least one of a duration of the media content, a code rate of the media content.

In some embodiments, the play status information includes: at least one of a duration of the remaining media content in the corresponding buffer of the currently played media content, a duration of the remaining media content in the corresponding buffer of the unreported media content, a number and duration of the jams that occur in the playing process of the currently played media content, a first frame time of the currently played media content, download information of the currently played media content, and download information of the unreported media content.

Referring now to fig. 9, which illustrates a model building apparatus provided by an embodiment of the present disclosure, the model building apparatus 600 includes:

The sample acquiring module 610 is configured to acquire a historical viewing information sample of a user, a play status information sample of an electronic device, and an attribute information sample of media content to be recommended.

The user experience index determining module 620 is configured to determine a user experience index of the user according to the play status information sample.

The training module 630 is configured to train the recommendation model with the user experience index as tag information of the historical viewing information sample, the playing state information sample, and the attribute information sample, so that the recommendation model determines the user experience index to be optimized based on the historical viewing information sample, and output recommendation result information for optimizing the user experience index according to the playing state information sample and the attribute information sample, where the recommendation result information is used to determine target media content to be preloaded from media content to be recommended.

In some embodiments, the sample acquisition module 610 includes:

And the simulation environment initialization sub-module is used for initializing the simulation environment based on preset network bandwidth data and real viewing behavior information of the user.

The sample acquisition sub-module is used for simulating the process of watching the media content by the user through the electronic equipment in the simulation environment, and acquiring historical watching information samples of the user, play state information samples of the electronic equipment and attribute information samples of the media content to be recommended, which are generated in the process.

In some embodiments, the network parameters of the recommendation model include a number of activations set by the upper layer model for each lower layer model, the number of activations being used to characterize a number of repetitions of media content recommendation by the recommendation model each time the lower layer model is employed, and the model construction apparatus 600 further includes:

And the user experience index updating module is used for acquiring new play state information samples again after each training, and determining new user experience indexes according to the new play state information samples.

And the network parameter updating module is used for updating the network parameters of the recommendation model according to the user experience index and the change condition of the new user experience index.

Referring now to fig. 10, a schematic diagram of an electronic device (e.g., a terminal device or server in fig. 1) 600 suitable for use in implementing embodiments of the present disclosure is shown. The terminal devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and the like, and stationary terminals such as digital TVs, desktop computers, and the like. The electronic device shown in fig. 10 is merely an example and should not be construed to limit the functionality and scope of use of the disclosed embodiments.

As shown in fig. 10, the electronic device 600 may include a processing means (e.g., a central processing unit, a graphic processor, etc.) 601, which may perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM) 602 or a program loaded from a storage means 608 into a Random Access Memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic apparatus 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. An input/output (I/O) interface 605 is also connected to bus 604.

In general, the following devices may be connected to the I/O interface 605: input devices 606 including, for example, a touch screen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, and the like; an output device 607 including, for example, a Liquid Crystal Display (LCD), a speaker, a vibrator, and the like; storage 608 including, for example, magnetic tape, hard disk, etc.; and a communication device 609. The communication means 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. While fig. 10 shows an electronic device 600 having various means, it is to be understood that not all of the illustrated means are required to be implemented or provided. More or fewer devices may be implemented or provided instead.

In particular, according to embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a non-transitory computer readable medium, the computer program comprising program code for performing the method shown in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via communication means 609, or from storage means 608, or from ROM 602. The above-described functions defined in the methods of the embodiments of the present disclosure are performed when the computer program is executed by the processing device 601.

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

In some embodiments, the clients, servers may communicate using any currently known or future developed network protocol, such as HTTP (HyperText Transfer Protocol ), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), the internet (e.g., the internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed networks.

The computer readable medium may be contained in the electronic device; or may exist alone without being incorporated into the electronic device.

The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to: acquiring historical viewing information of a user, play state information of electronic equipment and attribute information of media content to be recommended; inputting historical viewing information, play state information and attribute information of a user into a recommendation model, obtaining recommendation result information output by the recommendation model, determining user experience indexes to be optimized by the recommendation model based on the historical viewing information of the user, and outputting recommendation result information for optimizing the user experience indexes according to the play state information and the attribute information; and determining target media content from the media content to be recommended according to the recommendation result information to perform preloading.

Or the computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to: acquiring a historical viewing information sample of a user, a playing state information sample of electronic equipment and an attribute information sample of media content to be recommended; determining user experience indexes of the user according to the play state information samples; and training a recommendation model by taking the user experience index as tag information of the historical viewing information sample, the playing state information sample and the attribute information sample, so that the recommendation model determines the user experience index to be optimized based on the historical viewing information sample, and outputting recommendation result information for optimizing the user experience index according to the playing state information sample and the attribute information sample, wherein the recommendation result information is used for determining target media content to be preloaded from media content to be recommended.

Computer program code for carrying out operations of the present disclosure may be written in one or more programming languages, including, but not limited to, an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer (for example, through the Internet using an Internet service provider).

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

The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: a Field Programmable Gate Array (FPGA), an Application Specific Integrated Circuit (ASIC), an Application Specific Standard Product (ASSP), a system on a chip (SOC), a Complex Programmable Logic Device (CPLD), and the like.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The foregoing description is only of the preferred embodiments of the present disclosure and description of the principles of the technology being employed. It will be appreciated by persons skilled in the art that the scope of the disclosure referred to in this disclosure is not limited to the specific combinations of features described above, but also covers other embodiments which may be formed by any combination of features described above or equivalents thereof without departing from the spirit of the disclosure. Such as those described above, are mutually substituted with the technical features having similar functions disclosed in the present disclosure (but not limited thereto).

Moreover, although operations are depicted in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order. In certain circumstances, multitasking and parallel processing may be advantageous. Likewise, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the present disclosure. Certain features that are described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination.

Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are example forms of implementing the claims. The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.

Claims

1. A method of preloading media content, comprising:

acquiring historical viewing information of a user, play state information of electronic equipment and attribute information of media content to be recommended;

determining target media content from the media content to be recommended according to the recommendation result information to perform preloading;

The recommendation model comprises an upper model and a plurality of lower models, wherein the lower models are in one-to-one correspondence with a plurality of user experience indexes, and each lower model learns the influence relationship of play state information and attribute information of a media content sample on the user experience indexes corresponding to the lower models in advance;

Inputting the historical viewing information, the playing state information and the attribute information of the user into a recommendation model to obtain recommendation result information output by the recommendation model, wherein the method comprises the following steps:

inputting the historical viewing information into the upper layer model, and determining a target lower layer model from the plurality of lower layer models according to the output result of the upper layer model;

and inputting the play state information and the attribute information into the target lower layer model to obtain the recommendation result information output by the target lower layer model.

2. The method of claim 1, wherein the output of the upper model includes a number of activations corresponding to the target lower model, the method further comprising:

After the target media content is determined to be preloaded from the media content to be recommended according to the recommendation result information, the following operations are repeatedly executed for the target lower layer model until the execution times reach the activation times:

Re-acquiring the play state information of the electronic equipment and the attribute information of the media content to be recommended;

Inputting new play state information and new attribute information into the target lower layer model to obtain new recommendation result information;

3. The method according to claim 2, wherein the method further comprises:

And when the execution times reach the activation times, acquiring the historical viewing information of the user again, and returning to execute the step of inputting the historical viewing information into the upper model according to the new historical viewing information.

4. The method according to claim 1, wherein the target lower layer model is plural, the inputting the play status information and the attribute information into the target lower layer model, and obtaining the recommendation result information output by the target lower layer model, includes:

Inputting the play state information and the attribute information into each target lower model to obtain recommendation result information output by each target lower model;

the step of determining target media content from the media content to be recommended according to the recommendation result information for preloading comprises the following steps:

and determining the target media content recommended by each piece of recommendation result information from the media content to be recommended to load.

5. The method of claim 1, wherein the historical viewing information comprises play state information of historical media content viewed by the user and/or historical behavior information of the historical media content viewed by the user, and wherein the user experience metrics to be optimized are determined based on the play state information of the historical media content and/or the historical behavior information.

6. The method of claim 5, wherein the historical behavior information includes at least one of a viewing duration of the user, a duration of the viewed media content, an exit operation when a pause occurs in the viewed media content, and an exit operation in waiting for a first frame time of the media content.

7. The method according to any one of claims 1-6, wherein the attribute information of the media content to be recommended includes at least one of a duration of the media content and a code rate of the media content.

8. The method according to any one of claims 1-6, wherein the play status information includes:

At least one of a duration of the remaining media content in the corresponding buffer of the currently played media content, a duration of the remaining media content in the corresponding buffer of the unreported media content, a number and duration of the jams that occur in the playing process of the currently played media content, a first frame time of the currently played media content, download information of the currently played media content, and download information of the unreported media content.

9. A method of modeling, comprising:

training a recommendation model by taking the user experience index as tag information of the historical viewing information sample, the play state information sample and the attribute information sample, so that the recommendation model determines the user experience index to be optimized based on the historical viewing information sample, and outputting recommendation result information for optimizing the user experience index according to the play state information sample and the attribute information sample, wherein the recommendation result information is used for determining target media content to be preloaded from the media content to be recommended;

The recommendation model comprises an upper model and a plurality of lower models, wherein the lower models are in one-to-one correspondence with a plurality of user experience indexes, each lower model learns the influence relationship of play state information and attribute information of a media content sample on the user experience index corresponding to the lower model in advance, the upper model is used for learning to determine the user experience index to be optimized based on the historical viewing information sample, and one lower model is used for learning the influence relationship of the play state information sample and the attribute information sample on one user experience index.

10. The method according to claim 9, wherein obtaining a historical viewing information sample of the user, a play status information sample of the electronic device, and an attribute information sample of the media content to be recommended comprises:

initializing a simulation environment based on preset network bandwidth data and real viewing behavior information of the user;

And simulating the process of watching the media content by the user through the electronic equipment in the simulation environment, and acquiring a historical watching information sample of the user, a playing state information sample of the electronic equipment and an attribute information sample of the media content to be recommended, which are generated in the process.

11. The method of claim 9, wherein the network parameters of the recommendation model include a number of activations set by an upper layer model for each of the lower layer models, the number of activations being used to characterize a number of repetitions of media content recommendation by the recommendation model each time the lower layer model is employed, the method further comprising:

After each training, re-acquiring a new play state information sample, and determining a new user experience index according to the new play state information sample;

And updating the network parameters of the recommendation model according to the user experience index and the change condition of the new user experience index.

12. A media content preloading device, comprising:

The information acquisition module is used for acquiring historical viewing information of a user, playing state information of the electronic equipment and attribute information of media content to be recommended;

The preloading module is used for determining target media content from the media content to be recommended according to the recommendation result information to preload;

The recommendation result information acquisition module comprises:

An upper layer model activation sub-module, configured to input the historical viewing information into the upper layer model, and determine a target lower layer model from the plurality of lower layer models according to an output result of the upper layer model;

And the lower model activation submodule is used for inputting the play state information and the attribute information into the target lower model to obtain the recommended result information output by the target lower model.

13. A model building apparatus, comprising:

the sample acquisition module is used for acquiring a historical viewing information sample of a user, a play state information sample of the electronic equipment and an attribute information sample of media content to be recommended;

The training module is used for taking the user experience index as the tag information of the historical viewing information sample, the play state information sample and the attribute information sample, training a recommendation model, enabling the recommendation model to determine the user experience index to be optimized based on the historical viewing information sample, outputting recommendation result information for optimizing the user experience index according to the play state information sample and the attribute information sample, and determining target media content to be preloaded from the media content to be recommended;

14. A computer readable medium on which a computer program is stored, characterized in that the program, when being executed by a processing device, carries out the steps of the method according to any one of claims 1-8.

15. An electronic device, comprising:

a storage device having a computer program stored thereon;

Processing means for executing said computer program in said storage means to carry out the steps of the method according to any one of claims 1-8.

16. A computer readable medium on which a computer program is stored, characterized in that the program, when being executed by a processing device, carries out the steps of the method according to any one of claims 9-12.

17. An electronic device, comprising:

a storage device having a computer program stored thereon;

processing means for executing said computer program in said storage means to carry out the steps of the method according to any one of claims 9-12.