CN113837809B

CN113837809B - Medium information quality prediction method, device, electronic equipment and storage medium

Info

Publication number: CN113837809B
Application number: CN202111152739.0A
Authority: CN
Inventors: 李岩; 陈翼翼; 刘旭东
Original assignee: Beijing Dajia Internet Information Technology Co Ltd
Current assignee: Beijing Dajia Internet Information Technology Co Ltd
Priority date: 2021-09-29
Filing date: 2021-09-29
Publication date: 2024-05-03
Anticipated expiration: 2041-09-29
Also published as: CN113837809A

Abstract

The disclosure relates to a media information quality prediction method, a device, an electronic device and a storage medium, comprising: acquiring historical data information of media information to be predicted, wherein the historical data information comprises first delivery data of the media information to be predicted in a preset time range, at least one associated media information of the same media information publisher corresponding to the media information to be predicted, and/or second delivery data of each associated media information in the preset time range; performing quality prediction according to the media information to be predicted and the historical data information of the media information to be predicted to obtain predicted delivery data of the media information to be predicted; wherein the delivery data comprises data for characterizing the delivery quality of the media information. The embodiment of the disclosure can improve the prediction precision of the delivery quality of the media information.

Description

Medium information quality prediction method, device, electronic equipment and storage medium

Technical Field

The disclosure relates to the technical field of data processing, and in particular relates to a media information quality prediction method, a device, electronic equipment and a storage medium.

Background

The media information may include information distributed to the public at the platform, and may include video information or image information in particular. Before the media information is put in, the method has important significance in predicting the put quality of the media information. For example: under the condition that the medium information is the advertisement material, the recall and the delivery of the advertisement material with high quality can be performed by predicting the delivery quality of the advertisement material, thereby being beneficial to accurately delivering the advertisement material and improving the conversion efficiency of the advertisement material.

In the related art, in the case that the user side information cannot be obtained before the media information is delivered, the delivery quality of the media information can be predicted through the content information of the media information. For example: the quality of the media information delivery can be predicted by whether the picture is clear/contains characters/whether music/intonation is fast or slow, etc.

However, the quality of the media information in the delivery link is not only related to the content information of the media information itself, for example: factors such as industry difference and time change have great influence on data such as click rate and/or conversion rate of media information. Accordingly, in the related art, the quality of delivery of the medium information is predicted only by the content information of the medium information itself, which results in a lower accuracy of prediction of the quality of delivery of the medium information.

Disclosure of Invention

The disclosure provides a method, a device, an electronic device and a storage medium for predicting quality of media information, so as to at least solve the problem of low prediction accuracy of the quality of media information in the related art. The technical scheme of the present disclosure is as follows:

According to a first aspect of an embodiment of the present disclosure, there is provided a media information quality prediction method, including:

Acquiring historical data information of media information to be predicted, wherein the historical data information comprises first delivery data of the media information to be predicted in a preset time range, at least one associated media information of the same media information publisher corresponding to the media information to be predicted, and/or second delivery data of each associated media information in the preset time range;

Performing quality prediction according to the media information to be predicted and the historical data information of the media information to be predicted to obtain predicted delivery data of the media information to be predicted;

wherein the delivery data comprises data for characterizing the delivery quality of the media information.

In a possible implementation manner, the quality prediction is performed according to the media information to be predicted and the historical data information of the media information to be predicted through a quality prediction network, so as to obtain predicted delivery data of the media information to be predicted, wherein the quality prediction network comprises a feature extraction network and a prediction network;

the step of realizing the quality prediction according to the media information to be predicted and the historical data information of the media information to be predicted through a quality prediction network to obtain predicted delivery data of the media information to be predicted comprises the following steps:

extracting features of the media information to be predicted and historical data information of the media information to be predicted through the feature extraction network to obtain material features corresponding to the media information to be predicted;

And carrying out prediction processing on the material characteristics through the prediction network to obtain predicted delivery data of the medium information to be predicted.

In a possible implementation manner, the historical data information comprises first delivery data of the media information to be predicted within a preset time range, and the feature extraction network comprises a first network and a second network;

The feature extraction of the media information to be predicted and the historical data information of the media information to be predicted through the feature extraction network is performed to obtain material features corresponding to the media information to be predicted, and the method comprises the following steps:

extracting features of the media information to be predicted through the first network to obtain first content features;

Vector conversion processing is carried out on the first put data through the second network, and a first vector representation is obtained;

And obtaining the material characteristics of the media information to be predicted according to the first content characteristics and the first vector representation.

In a possible implementation manner, the historical data information further includes at least one associated media information corresponding to the same media information publisher as the media information to be predicted, and second delivery data of each associated media information within the preset time range;

before the obtaining the material characteristics of the media information to be predicted according to the first content characteristics and the first vector representation, the method further includes:

Extracting the characteristics of each piece of associated media information through the first network to obtain each second content characteristic;

vector conversion processing is carried out on each second put-in data through the second network respectively, so that each second vector representation is obtained;

And obtaining the material characteristics of the media information to be predicted according to the first content characteristics and the first vector representation, wherein the material characteristics comprise:

obtaining a first material characteristic of the media information to be predicted according to the first content characteristic and the first vector representation;

Obtaining second material characteristics of the media information to be predicted according to the second content characteristics and the second vector representations;

and obtaining the material characteristics of the medium information to be predicted according to the first material characteristics and the second material characteristics.

In a possible implementation manner, the feature extraction network further includes a fusion network, and the obtaining the material feature of the media information to be predicted according to the first material feature and the second material feature includes:

And carrying out feature fusion processing on the first material features and the second material features through the fusion network to obtain the material features of the medium information to be predicted.

In one possible implementation manner, the first network includes at least one of a first feature extraction module, a second feature extraction module, a third feature extraction module, and a fourth feature extraction module;

the first feature extraction module is used for extracting visual features of the medium information;

the second feature extraction module is used for extracting image text features of the medium information;

the third feature extraction module is used for extracting voice text features of the medium information;

the fourth feature extraction module is used for extracting the audio features of the media information.

In one possible implementation manner, before the quality prediction is performed according to the media information to be predicted and the historical data information of the media information to be predicted through the quality prediction network, the method further includes:

Training the quality prediction network by adopting a pre-constructed training set, wherein the training set comprises a plurality of sample groups, and the sample groups comprise sample medium information, historical data information of the sample medium information and labeling and throwing data of the sample medium information;

The training of the quality prediction network using the pre-constructed training set includes:

The sample medium information and the historical data information of the sample medium information are subjected to prediction processing through a quality prediction network, so that predicted delivery data corresponding to the sample medium information are obtained;

according to the predicted delivery data corresponding to the sample medium information and the marked delivery data of the sample medium information, determining the predicted loss of the quality prediction network;

training the quality prediction network based on the prediction loss.

According to a second aspect of the embodiments of the present disclosure, there is provided a media information quality prediction apparatus including:

An acquisition unit configured to perform acquisition of history data information of media information to be predicted, the history data information including first delivery data of the media information to be predicted within a preset time range, at least one associated media information corresponding to the same media information publisher as the media information to be predicted, and/or second delivery data of each of the associated media information within the preset time range;

the prediction unit is configured to perform quality prediction according to the media information to be predicted and the historical data information of the media information to be predicted, so as to obtain predicted delivery data of the media information to be predicted;

The prediction unit includes:

the first feature extraction subunit is configured to perform feature extraction on the media information to be predicted and the historical data information of the media information to be predicted through the feature extraction network to obtain material features corresponding to the media information to be predicted;

the first prediction subunit is configured to execute prediction processing on the material characteristics through the prediction network to obtain predicted delivery data of the medium information to be predicted.

In one possible implementation manner, the historical data information includes first delivery data of the media information to be predicted within a preset time range, the feature extraction network includes a first network and a second network, and the first feature extraction subunit includes:

the feature extraction sub-module is configured to perform feature extraction on the media information to be predicted through the first network to obtain first content features;

The vector conversion sub-module is configured to perform vector conversion processing on the first delivery data through the second network to obtain a first vector representation;

And the processing sub-module is configured to execute the material characteristics of the media information to be predicted according to the first content characteristics and the first vector representation.

In a possible implementation manner, the historical data information further includes at least one associated media information corresponding to the same media information publisher as the media information to be predicted, and second delivery data of each associated media information within the preset time range; the medium information quality prediction apparatus further includes:

A second feature extraction unit configured to perform feature extraction on each of the associated media information through the first network, respectively, to obtain each of second content features;

The second vector conversion unit is configured to perform vector conversion processing on each piece of second delivery data through the second network to obtain each second vector representation;

the processing sub-module is further configured to perform:

In one possible implementation, the feature extraction network further includes a fusion network, and the processing sub-module is further configured to perform:

In one possible implementation manner, the media information quality prediction apparatus further includes:

A training unit configured to perform training of the quality prediction network using a pre-constructed training set, the training set including a plurality of sample groups including sample media information, historical data information of the sample media information, and annotation delivery data of the sample media information;

The training unit is further configured to perform:

training the quality prediction network based on the prediction loss.

According to a third aspect of embodiments of the present disclosure, there is provided an electronic device, comprising:

A processor;

a memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the media information quality prediction method of any of the preceding claims.

According to a fourth aspect of embodiments of the present disclosure, there is provided a computer readable storage medium, which when executed by a processor of an electronic device, causes the electronic device to perform the medium information quality prediction method as set forth in any one of the preceding claims.

According to a fifth aspect of embodiments of the present disclosure, there is provided a computer program product comprising instructions therein, which when executed by a processor of an electronic device, enable the electronic device to perform the medium information quality prediction method as set forth in any one of the preceding claims.

The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:

According to the embodiment of the disclosure, the predicted delivery data of the medium information to be predicted is obtained by acquiring the historical data information of the medium information to be predicted and carrying out quality prediction according to the medium information to be predicted and the historical data information of the medium information to be predicted. The historical data information comprises first throwing data of the media information to be predicted in a preset time range, at least one associated media information corresponding to the same media information publisher of the media information to be predicted, and/or second throwing data of each associated media information in the preset time range. According to the media information quality prediction method, device, electronic equipment and storage medium provided by the embodiment of the disclosure, under the condition that the user side information cannot be obtained, the quality of the media information to be predicted can be predicted by introducing the historical data information of the media information to be predicted.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure and do not constitute an undue limitation on the disclosure.

Fig. 1 is a flowchart illustrating a media information quality prediction method according to an exemplary embodiment.

Fig. 2 is a block diagram of a quality prediction network, according to an example embodiment.

Fig. 3 is a flowchart illustrating a media information quality prediction method according to an exemplary embodiment.

Fig. 4 is a flowchart illustrating a media information quality prediction method according to an exemplary embodiment.

Fig. 5 is a flowchart illustrating a media information quality prediction method according to an exemplary embodiment.

Fig. 6 is a schematic diagram illustrating a media information quality prediction method according to an exemplary embodiment.

Fig. 7 is a flowchart illustrating a media information quality prediction method according to an exemplary embodiment.

Fig. 8 is a block diagram illustrating a quality prediction network in accordance with an exemplary embodiment.

Fig. 9 is a block diagram of a converged network in accordance with an exemplary embodiment.

Fig. 10 is a flowchart illustrating a media information quality prediction method according to an exemplary embodiment.

Fig. 11 is a block diagram illustrating a media information quality prediction apparatus according to an exemplary embodiment.

Fig. 12 is a block diagram of an electronic device, according to an example embodiment.

Fig. 13 is a block diagram of an electronic device, according to an example embodiment.

Detailed Description

In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the foregoing figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the disclosure described herein may be capable of operation in sequences other than those illustrated or described herein. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.

It should be further noted that, the user information (including, but not limited to, user equipment information, user personal information, etc.) and the data (including, but not limited to, data for presentation, analyzed data, etc.) related to the present disclosure are information and data authorized by the user or sufficiently authorized by each party.

In one embodiment, as shown in fig. 1, a method for predicting the quality of media information is provided, where this embodiment is applied to a terminal for illustration, it is understood that the method may also be applied to a server, and may also be applied to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the steps of:

in step 102, historical data information of the media information to be predicted is obtained, where the historical data information includes first delivery data of the media information to be predicted within a preset time range, at least one associated media information corresponding to the same media information publisher as the media information to be predicted, and/or second delivery data of each associated media information within the preset time range.

In the embodiment of the disclosure, the media information to be predicted is media information of the quality to be predicted. The media information base may store media information released by each media information publisher and relevant release data released by each media information history, and in the case where the media information is an advertisement material, the media information publisher may be an advertiser releasing the advertisement material. By way of example, the delivery data may include data capable of characterizing the delivery quality of the media information, including, but not limited to, information such as the number of clicks, the number of exposures, the number of conversions, or data information determined by at least two of the number of clicks, the number of exposures, the number of conversions, etc., such as: the ratio determined by the conversion number to the exposure number can be identified as ctcvr. The embodiment of the disclosure does not specifically limit the delivery data, and can be set by a user according to a predicted requirement, and the embodiment of the disclosure is described below by taking a ratio ctcvr of the delivery data as a conversion number to an exposure number as an example, and when the delivery data is the ratio of the conversion number to the exposure number, the higher the ratio is, the better the delivery quality of the medium information is represented.

After determining the media information to be predicted, historical data information of the media information to be predicted can be obtained from a media information base, wherein the historical data information can comprise first delivery data generated in a preset time range, at least one associated media information corresponding to the same media information publisher of the media information to be predicted, and/or second delivery data generated in the preset time range by each associated media information in the historical delivery process of the media information to be predicted.

The preset time range is a preset time range, and the specific value can be determined by a user according to the predicted requirement. For example: the time range can be set to be within one week, the corresponding delivery data of the media information to be predicted within one week can be obtained from the media information base, and the delivery data (ctcvr 1, ctcvr2, ctcvr3, ctcvr, ctcvr5, ctcvr, ctcvr 7) of the media information to be predicted within one week can be obtained assuming that the delivery data is a ratio ctcvr of conversion number to exposure number. And the method can also acquire at least one piece of other media information released by the media information publisher from the media information base as the associated media information of the media information to be predicted according to the media information publisher corresponding to the media information to be predicted, and acquire the second release data of each associated media information in a week from the media information base.

In step 104, quality prediction is performed according to the media information to be predicted and the historical data information of the media information to be predicted, so as to obtain predicted delivery data of the media information to be predicted; wherein the delivery data includes data characterizing the quality of delivery of the media information.

In the embodiment of the disclosure, after obtaining the media information to be predicted and the historical data information of the media information to be predicted, quality prediction may be performed according to the media information to be predicted and the historical data information of the media information to be predicted, so as to obtain predicted delivery data of the media information to be predicted. For example: the first predicted delivery data of the medium information to be predicted may be determined in combination with the content included in the medium information to be predicted, the second predicted delivery data of the medium information to be predicted may be determined according to the history data information (the second predicted delivery data may be determined according to the first delivery data and/or the second delivery data, for example, a mean value of the first delivery data and/or the second delivery data may be used as the second predicted delivery data), and the first predicted delivery data and the second predicted delivery data may be weighted and summed to obtain the predicted delivery data of the medium information to be predicted.

Or the predicted delivery data of the medium information to be predicted can be obtained by performing prediction processing on the medium information to be predicted and the historical data information of the medium information to be predicted through a pre-trained neural network.

After obtaining the predicted delivery data of the media information to be predicted, the media information to be predicted can be recalled and link ordered according to the predicted delivery data, for example: and under the condition that the predicted delivery data is larger than the delivery threshold value, the media information to be predicted is recalled, and the position sequence of the media information to be predicted in the recall link is determined according to the predicted delivery data, so that the high-quality media information can enter the recall link, and the benefits of a media information publisher and a platform are improved.

According to the embodiment of the disclosure, the predicted delivery data of the medium information to be predicted is obtained by acquiring the historical data information of the medium information to be predicted and carrying out quality prediction according to the medium information to be predicted and the historical data information of the medium information to be predicted. The historical data information comprises first throwing data of the media information to be predicted in a preset time range, at least one associated media information corresponding to the same media information publisher of the media information to be predicted, and/or second throwing data of each associated media information in the preset time range. According to the media information quality prediction method provided by the embodiment of the disclosure, under the condition that the user side information cannot be obtained, the release quality of the media information to be predicted can be predicted by introducing the historical data information of the media information to be predicted, and as the historical data information can reflect potential favorites of most users to a certain extent, the release quality of the media information to be predicted is predicted by combining the historical data information of the media information to be predicted, and the prediction precision of the release quality can be improved.

In an exemplary embodiment, quality prediction according to the media information to be predicted and the historical data information of the media information to be predicted may be implemented through a quality prediction network, so as to obtain predicted delivery data of the media information to be predicted, as shown in fig. 2, where the quality prediction network includes a feature extraction network and a prediction network.

As shown in fig. 3, the quality prediction is implemented by the quality prediction network according to the media information to be predicted and the historical data information of the media information to be predicted, specifically, the method may be implemented by the following steps:

In step 302, feature extraction is performed on the media information to be predicted and the historical data information of the media information to be predicted through a feature extraction network, so as to obtain material features corresponding to the media information to be predicted;

in step 304, the material characteristics are predicted by the prediction network, so as to obtain predicted delivery data of the media information to be predicted.

In an embodiment of the present disclosure, the quality prediction network includes a feature extraction network and a prediction network. The feature extraction network is used for extracting material features, and the prediction network is used for predicting according to the material features to obtain predicted delivery data. After the media information to be predicted and the historical data information corresponding to the media information to be predicted are obtained, the media information to be predicted and the historical data information of the media information to be predicted can be input into a feature extraction network in a quality prediction network as input information to perform feature extraction, and after the material features of the media information to be predicted are obtained, the material features of the media information to be predicted are input into the prediction network as input information to perform prediction processing, so that predicted delivery data of the media information to be predicted are obtained.

For example: the historical data information comprises a plurality of associated media information of the same advertiser corresponding to the media information to be predicted, the media information to be predicted and the associated media information can be input into the feature extraction network to perform feature extraction to obtain material features of the media information to be predicted, and the material features are input into the prediction network to perform prediction processing to obtain predicted delivery data of the media information to be predicted.

According to the media information quality prediction method provided by the embodiment of the disclosure, the prediction of the delivery quality of the media information to be predicted according to the media information to be predicted and the historical data information of the media information to be predicted can be realized through the quality prediction network, namely, the characteristics of the historical data information are integrated into the characteristics of the media information to be predicted, so that the preference of a user is known, and the prediction efficiency and the prediction precision of the delivery quality can be improved.

In an exemplary embodiment, before implementing quality prediction according to the media information to be predicted and the historical data information of the media information to be predicted through the quality prediction network to obtain the predicted delivery data of the media information to be predicted, the method may further include: and training the quality prediction network by adopting a pre-constructed training set, wherein the training set comprises a plurality of sample groups, and the sample groups comprise sample medium information, historical data information of the sample medium information and labeling put-in data of the sample medium information. Referring to fig. 4, training a quality prediction network using a pre-constructed training set may be specifically implemented by:

in step 402, a quality prediction network predicts sample media information and historical data information of the sample media information to obtain predicted delivery data corresponding to the sample media information;

In step 404, according to the predicted delivery data corresponding to the sample media information and the labeling delivery data of the sample media information, determining a predicted loss of the quality prediction network;

In step 406, the quality prediction network is trained based on the predicted loss.

In the embodiment of the disclosure, a sample group can be constructed according to the medium information in the medium information base as sample medium information, the history data information of the medium information as the history data information of the sample medium information, and the current delivery data of the medium information as the label delivery data of the sample medium information, a training set can be constructed according to the sample medium information, the history data information of the sample medium information, and the label delivery data of the sample medium information, and a training set can be constructed according to a plurality of sample groups, and a network can be predicted according to the training quality of the training set.

For example, the sample advertisement data and the historical data information of the sample advertisement data can be used as input information to be input into a quality prediction network for quality prediction, and the output of the quality prediction network is the predicted delivery data corresponding to the sample medium information. After the predicted delivery data of the sample medium information is obtained, the predicted loss of the quality prediction network can be determined according to the predicted delivery data of the sample medium information and the marked delivery data of the sample medium information. The embodiment of the present disclosure does not specifically limit the specific manner of determining the predicted loss, and virtually any loss calculation manner is applicable to the embodiment of the present disclosure, for example: the loss calculation modes such as MSE (Mean Square Error ) and MAE (Mean Absolute Error, absolute error).

It should be noted that, in the embodiment of the present disclosure, the network structure of the predicted network is not specifically limited, for example: the predictive network may be a DNN (Deep Neural Networks, deep neural network) network comprising multiple fully connected layers and relu activation functions, etc. modules.

After determining the predicted loss of the quality prediction network, the quality prediction network may be trained based on the predicted loss. For example, in the case where the predicted loss does not meet the training requirement (e.g., the predicted loss is greater than a preset loss threshold), the network parameters of the quality prediction network may be adjusted according to the predicted loss, including adjusting the parameters of the feature extraction network and adjusting the parameters of the quality prediction network, until the predicted loss of the quality prediction network meets the training requirement (e.g., the predicted loss is less than or equal to the preset loss threshold), to obtain the trained quality prediction network.

According to the media information quality prediction method provided by the embodiment of the disclosure, the training set can be constructed according to the sample media information and the historical data information of the sample media information, and the quality prediction network is trained through the constructed training set, so that the prediction of the delivery quality of the media information to be predicted according to the media information to be predicted and the historical data information of the media information to be predicted can be realized through the quality prediction network, and the prediction efficiency and the prediction precision of the delivery quality can be improved.

In an exemplary embodiment, the historical data information includes first delivery data of the media information to be predicted within a preset time range, and the feature extraction network includes a first network and a second network. Referring to fig. 5, in step 302, feature extraction is performed on media information to be predicted and historical data information of the media information to be predicted through a feature extraction network, so as to obtain material features corresponding to the media information to be predicted, which may be specifically implemented by the following steps:

in step 502, extracting features of media information to be predicted through a first network to obtain first content features;

In step 504, vector conversion processing is performed on the first delivery data through the second network, so as to obtain a first vector representation;

In step 506, material characteristics of the media information to be predicted are obtained according to the first content characteristics and the first vector representation.

In the embodiment of the disclosure, referring to fig. 6, the historical data information includes first delivery data of the media information to be predicted within a preset time range, the feature extraction network includes a first network and a second network, where the first network is used for feature extraction of the network to be predicted to obtain a first content feature, and the second network is used for vector conversion processing of the first delivery data of the network to be predicted to obtain a first vector representation of the first delivery data.

In one example, the first network may include at least one of a first feature extraction module for extracting visual features of the media information, a second feature extraction module for extracting image text features of the media information, a third feature extraction module for extracting phonetic text features of the media information, and a fourth feature extraction module for extracting audio features of the media information.

The first feature extraction module, the second feature extraction module, the third feature extraction module, and the fourth feature extraction module may be pre-trained network modules. In the embodiment of the present disclosure, the network structures of the first feature extraction module, the second feature extraction module, the third feature extraction module, and the fourth feature extraction module are not limited, any network module capable of implementing the corresponding feature extraction function is suitable for use in the embodiment of the present disclosure, and for convenience of understanding, in the embodiment of the present disclosure, the first feature extraction module is a moco (Momentum Contrast for Unsupervised Visual Representation Learning, non-supervised visual representation learning based on momentum comparison) model, the second feature extraction module and the third feature extraction module are a Bert (Bidirectional Encoder Representations from Transformers, bi-directional coding representation based on a converter) model, and the fourth feature extraction module is a vggish model.

The first feature extraction module may employ different visual feature extraction strategies according to different types of media information. For example, for the media information of the picture type, the first feature extraction module may directly extract moco features of the picture as visual features of the media information; for the media information of the video type, the first feature extraction module may average moco features of the multi-frame image (for example, extract 8 frames) from the video, and then average moco features of the multi-frame image to obtain the visual features of the media information of the video type.

The second feature extraction module may also employ different visual feature extraction strategies according to different types of media information. For the media information of the picture type, text information appearing in the picture/video can be extracted by using methods such as OCR (Optical Character Recognition ) and the like, and then the image text characteristics in the text information are extracted; for the media information of the video type, extracting multiple frames of images (for example, extracting 8 frames) from the video, extracting text information for each frame of images, splicing the text information of the multiple frames of images, and extracting the image text characteristics of the spliced text information.

The third feature extraction module is similar to the second feature extraction module, and extracts the voice text features in the voice text information after extracting the voice text information in the video by using ASR (Automatic Speech Recognition, automatic voice recognition technology). The fourth feature extraction module may directly perform audio feature extraction on audio data in the video.

After the characteristics are extracted through the multiple items in the first characteristic extraction module, the second characteristic extraction module, the third characteristic extraction module and the fourth characteristic extraction module, the extracted multiple items of characteristics can be spliced to obtain the first content characteristics of the media information to be predicted.

The second network is used for vector conversion of the put data, the network structure of the second network is not limited in the embodiment of the disclosure, and any network structure capable of realizing vector expression is in the application range of the embodiment of the disclosure.

Illustratively, the second network may be modeled based on an anchor interpolation concept. The second network is used for vector conversion of the input data, taking the ratio ctcvr of the input data as a conversion number and an exposure number as an example, ctcvr is a random number within a value range of 0-1, and has infinite values, and the second network can convert the input data into a fixed-dimension expression through anchor point interpolation. By way of example, a plurality ctcvr of anchor points may be pre-designed, for example: 8 total ctcvr of [1e-7,1e-6,1e-5,1e-4,1e-3,1e-2,1e-1,1] were determined to be 8 anchors. For a plurality of ctcvr anchor points, taking the logarithm with the base of 10 to obtain the corresponding numerical value of each anchor point, taking the above 8 anchor points as an example, taking the logarithm with the base of 10 to obtain [ -7, -6, -5, -4, -3, -2, -1,0].

When the second network is constructed, a vector representation Emb _i may be randomly initialized for each of the plurality of anchors, where Emb _i represents a vector representation of an anchor with a value i, and the second network is constructed according to the vector representations Emb _i of the plurality of anchors. For example, after the 8 anchor points are randomly initialized to be represented by vectors, [Emb_-7,Emb_-6,Emb_-5,Emb_-4,Emb_-3,Emb_-2,Emb_-1,Emb₀], may be obtained to construct a second network according to the 8 anchor points. And in the training process of the quality prediction network, vector representations of a plurality of anchor points of the second network are continuously adjusted until vector representations corresponding to the anchor points are obtained after the training is completed.

For the first delivery data, a logarithm based on 10 may be taken for the first delivery data to obtain a corresponding numerical value, two anchor points (assumed to be a first anchor point and a second anchor point) adjacent to the numerical value are determined, and the vector representations of the two adjacent anchor points are linearly combined to obtain a first vector representation of the first delivery data, and the specific process may refer to the following formula (1).

Wherein the Emb _ctcvr is used for representing a first vector representation corresponding to the first delivery data,Vector representation for representing a first anchor point,/>Vector representation for representing the second anchor point ctcvr represents the first delivery data, ctcvr _j-1 represents the delivery data corresponding to the first anchor point, and ctcvr _j represents the delivery data corresponding to the second anchor point.

Taking the first delivery data as 0.005 as an example, the second network takes the logarithm based on 10 to obtain-2.30, and the obtained first vector represents the reference formula (2) because-2.30 is positioned between anchor points corresponding to the-3 value and the-2 value:

Emb _-2.3＝(-2-(-2.3))*Emb_-3+(-2.3-(-3))*Emb_-2 formula (2)

It should be noted that, the above-mentioned setting of 8 anchor points is only an example of the embodiment of the disclosure, in fact, N anchor points may be set, and the value of N may be set according to the prediction requirement. When first delivery data smaller than the corresponding minimum delivery data in the plurality of anchor points appears, the vector representation of the anchor point with the minimum corresponding delivery data can be used as the vector representation of the first delivery data; accordingly, when first delivery data larger than corresponding maximum delivery data among the plurality of anchor points occurs, the vector representation of the anchor point with the corresponding maximum delivery data may be used as the vector representation of the first delivery data. Taking the above 8 anchor points as an example, when the first delivery data is smaller than 1e-7, the vector representation corresponding to the anchor point (1 e-7) can be used as the vector representation of the first delivery data; or when the first delivery data is greater than 1, the vector representation corresponding to the anchor point (1) can be used as the vector representation of the first delivery data.

Therefore, the second network can realize vector representation of any one of the put data by using vector representations of a plurality of anchor points, the second network is lighter, and the prediction efficiency of the quality prediction network can be improved.

After the first content characteristics and the first vector representations are obtained, the first content characteristics and the first vector representations can be spliced to obtain the material characteristics of the media information to be predicted, and further, the material characteristics are predicted through a prediction network, so that the predicted delivery data of the media information to be predicted can be obtained.

In the embodiment of the present disclosure, the first network may be a pre-trained network, that is, the first network is not trained in the training process of the quality prediction network; or the first network may also perform synchronous training with the second network and the prediction network during the training process of the quality prediction network, which is not specifically limited in the embodiments of the present disclosure.

According to the media information quality prediction method provided by the embodiment of the disclosure, under the condition that the user side information cannot be obtained, the quality of the media information to be predicted can be predicted by introducing the first delivery data of the media information to be predicted within the preset time range, and as the first delivery data of the media information to be predicted within the preset time range can reflect the potential preference of most users to a certain extent, the delivery quality of the media information to be predicted can be predicted by combining the first delivery data of the media information to be predicted within the preset time range, and the prediction precision of the delivery quality can be improved.

In an exemplary embodiment, the history data information may further include at least one associated media information of the same media information publisher corresponding to the media information to be predicted, and second delivery data of each associated media information within a preset time range. Referring to fig. 7, before obtaining the material characteristic of the media information to be predicted according to the first content characteristic and the first vector representation in step 506, the method may further include:

In step 508, feature extraction is performed on each associated media information through the first network, so as to obtain each second content feature;

in step 510, vector conversion processing is performed on each second delivery data through the second network, so as to obtain each second vector representation;

in step 506, according to the first content feature and the first vector representation, the material feature of the media information to be predicted is obtained, which may be specifically implemented by the following steps:

in step 5062, obtaining a first material feature of the media information to be predicted according to the first content feature and the first vector representation;

in step 5064, obtaining second material characteristics of the media information to be predicted according to the second content characteristics and the second vector representations;

in step 5066, the material characteristics of the media information to be predicted are obtained according to the first material characteristics and the second material characteristics.

In the embodiment of the disclosure, the historical data information includes first delivery data of the media information to be predicted within a preset time range, at least one associated media information of the same media information publisher corresponding to the media information to be predicted, and second delivery data of each associated media information within the preset time range.

On the basis that the first content characteristics and the first vector representation of the media information to be predicted are obtained in the foregoing embodiment, the first network may be used to extract characteristics of each associated media information to obtain second content characteristics of each associated media information, and the second network may be used to perform vector conversion on the second delivery data of each associated media information to obtain second vector representations of each second delivery data.

And performing splicing processing on the first content characteristic Emb _v and the first vector representation Emb _ctcvr to obtain a first material characteristic Emb1= (Emb _v;Emb_ctcvr) of the media information to be predicted. And respectively performing splicing processing on the second content features and the second vector representations corresponding to the associated media information to obtain second material features [ Emb21, emb22, … …, emb2n ] corresponding to the media information to be predicted, wherein Emb2 i= (Emb _vi;Emb_ctcvri), i is a positive number larger than 0 and smaller than n, n is the total number of the associated media information, emb2i represents the material features corresponding to the i associated media information, emb _vi represents the second content features corresponding to the i associated media information, and Emb _ctcvri represents the second vector representation corresponding to the i associated media information.

After the first material characteristic and the second material characteristic of the medium information to be predicted are obtained, fusion processing can be carried out on the first material characteristic and the second material characteristic, and the material characteristic of the medium information to be predicted is obtained. And further, the material characteristics can be predicted through a prediction network to obtain predicted delivery data of the media information to be predicted.

According to the media information quality prediction method provided by the embodiment of the disclosure, under the condition that user side information cannot be obtained, the quality of the media information to be predicted can be predicted by introducing the first delivery data of the media information to be predicted in the preset time range, at least one associated media information of the same media information publisher corresponding to the media information to be predicted, and the second delivery data of each associated media information in the preset time range, and the potential preference of most users can be accurately reflected through rich historical data information, so that the quality of the media information to be predicted can be predicted by combining the historical data information, and the prediction accuracy of the quality of the media information to be predicted can be improved.

In an exemplary embodiment, referring to fig. 8, the feature extraction network further includes a fusion network, and in step 5126, the material features of the media information to be predicted are obtained according to the first material features and the second material features, which may be specifically implemented as follows:

And carrying out feature fusion processing on the first material features and the second material features through a fusion network to obtain the material features of the medium information to be predicted.

In the embodiment of the present disclosure, the fusion network is used to perform feature fusion on the first material feature and the second material feature, and the embodiment of the present disclosure does not specifically limit the network structure of the fusion network, and any network capable of implementing feature fusion is suitable for the embodiment of the present disclosure.

Illustratively, referring to FIG. 9, a converged network can be constructed according to self-attention in a transducer framework. The fusion network can project the first material features through linear projection to reduce the feature dimension of the first material features, so as to obtain Q (query) corresponding to the first material features, and if the projected feature expression dimension is D, the feature dimension of Q is 1 xD at the moment. Similarly, the second material feature can be projected through linear projection to obtain K (key) and V (value) corresponding to the second material feature, after projection, the feature expression dimension of the second material feature is D the same as that of the first material feature, and after linear projection, the feature dimension of K, V is n×D.

The similarity between Q, K and the N pieces of associated media information may be determined, and the similarity may be used to characterize the similarity between the media information to be predicted and the N pieces of associated media information, and if the similarity between the media information to be predicted and a certain piece of associated media information is good, the similarity between Q, K may be infinitely close to 1, whereas if the similarity between the media information to be predicted and a certain piece of associated media information is poor, the similarity between Q, K may be extremely low, and may be infinitely close to 0.

After normalization processing is performed on each similarity, each similarity can be used as a weight, weighted summation is performed on each similarity and the corresponding V, so that fusion of the first material characteristic and the second material characteristic is achieved, and the material characteristic of the medium information to be predicted is obtained, and specifically, the following formula (3) can be referred to.

Wherein, the Attention (Q, K, V) is used for representing the corresponding predicted delivery data of the media information to be predicted. Q is used to represent queries, K is used to represent keys, V is used to represent values, and D is used to represent feature expression dimensions.

After the material characteristics are input into a prediction network for prediction processing, prediction delivery data corresponding to the media information to be predicted can be obtained, and then advertisement recall is carried out on the media information to be predicted according to the prediction delivery data corresponding to the media information to be predicted.

It should be noted that, the converged network may perform synchronous training with the first network, the second network, the prediction network, and other networks in the training process of the quality prediction network, that is, in the process of adjusting the network parameters of the quality prediction network according to the prediction loss, the network parameters of the converged network are adjusted until the prediction loss meets the training requirement, so as to obtain a trained converged network.

According to the media information quality prediction method provided by the embodiment of the disclosure, the first material characteristics and the second material characteristics can be fused through the fusion network, so that the material characteristics including potential favorites of the user are obtained, and further, quality prediction is performed according to the material characteristics, and the prediction precision of the delivery quality can be improved.

In order for those skilled in the art to better understand the disclosed embodiments, the disclosed embodiments are described below by way of specific examples.

Referring to fig. 10, the quality prediction network includes a feature extraction network and a prediction network, where the feature extraction network includes a first network, a second network, and a fusion network, and in this example, the media information to be predicted is taken as a video type, and the first network includes a first feature extraction module, a second feature extraction module, a third feature extraction module, and a fourth feature extraction module.

The method comprises the steps of obtaining historical data information corresponding to media information to be predicted, wherein the historical data information comprises a first ctcvr (ratio of conversion number to exposure number) of the media information to be predicted in a preset time range, and at least one associated media information of the same media information publisher corresponding to the media information to be predicted and a second ctcvr corresponding to the associated media information.

The media information to be predicted, the first ctcvr, at least one associated media information, and the second ctcvr of each associated media information are input into a quality prediction network. The first content features of the media information to be predicted and the second content features of the associated media information are extracted through the first network respectively (the specific process refers to the foregoing embodiment, and the description of this example is omitted here). The first ctcvr of the media information to be predicted is converted to a first vector representation and the second ctcvr of each associated media information is converted to a second vector representation, respectively, over a second network.

And in the fusion network, splicing the first content features and the first vector representation into first material features, splicing the second content features and the second vector representation into second material features respectively, fusing the first material features and the second material features, and inputting the material features into the prediction network after obtaining the material features of the medium information to be predicted.

And the prediction network predicts the material characteristics to obtain the prediction ctcvr of the medium information to be predicted.

It should be understood that, although the steps in the flowcharts of fig. 1 to 10 are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least a portion of the steps of fig. 1-10 may include multiple steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the steps or stages are performed necessarily occur sequentially, but may be performed alternately or alternately with other steps or at least a portion of the steps or stages in other steps.

It should be understood that the same/similar parts of the embodiments of the method described above in this specification may be referred to each other, and each embodiment focuses on differences from other embodiments, and references to descriptions of other method embodiments are only needed.

Fig. 11 is a block diagram illustrating a media information quality prediction apparatus according to an exemplary embodiment. Referring to fig. 11, the apparatus includes an acquisition unit 1102 and a prediction unit 1104.

The acquiring unit 1102 is configured to perform acquiring historical data information of media information to be predicted, where the historical data information includes first delivery data of the media information to be predicted in a preset time range, at least one associated media information corresponding to the same media information publisher as the media information to be predicted, and/or second delivery data of each associated media information in the preset time range;

a prediction unit 1104 configured to perform quality prediction according to the media information to be predicted and the historical data information of the media information to be predicted, so as to obtain predicted delivery data of the media information to be predicted;

According to the embodiment of the disclosure, the predicted delivery data of the medium information to be predicted is obtained by acquiring the historical data information of the medium information to be predicted and carrying out quality prediction according to the medium information to be predicted and the historical data information of the medium information to be predicted. The historical data information comprises first throwing data of the media information to be predicted in a preset time range, at least one associated media information corresponding to the same media information publisher of the media information to be predicted, and/or second throwing data of each associated media information in the preset time range. According to the media information quality prediction device provided by the embodiment of the disclosure, under the condition that the user side information cannot be obtained, the release quality of the media information to be predicted can be predicted by introducing the historical data information of the media information to be predicted, and as the historical data information can reflect the potential preference of most users to a certain extent, the release quality of the media information to be predicted is predicted by combining the historical data information of the media information to be predicted, and the prediction precision of the release quality can be improved.

In an exemplary embodiment, the quality prediction is performed according to the media information to be predicted and the historical data information of the media information to be predicted through a quality prediction network, so as to obtain predicted delivery data of the media information to be predicted, wherein the quality prediction network comprises a feature extraction network and a prediction network; the prediction unit 1104 includes:

In an exemplary embodiment, the historical data information includes first delivery data of the media information to be predicted within a preset time range, the feature extraction network includes a first network and a second network, and the first feature extraction subunit includes:

In an exemplary embodiment, the historical data information further includes at least one associated media information of the same advertiser corresponding to the media information to be predicted, and second delivery data of each of the associated media information within the preset time range; the medium information quality prediction apparatus further includes:

the processing sub-module is further configured to perform:

In an exemplary embodiment, the feature extraction network further comprises a fusion network, the processing sub-module further configured to perform:

In an exemplary embodiment, the first network includes at least one of a first feature extraction module, a second feature extraction module, a third feature extraction module, and a fourth feature extraction module;

In an exemplary embodiment, the media information quality prediction apparatus further includes:

The training unit is further configured to perform:

training the quality prediction network based on the prediction loss.

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.

Fig. 12 is a block diagram illustrating an electronic device 1200 for media information quality prediction, according to an example embodiment. For example, the electronic device 1200 may be a mobile phone, computer, digital broadcast terminal, messaging device, game console, tablet device, medical device, exercise device, personal digital assistant, or the like.

Referring to fig. 12, an electronic device 1200 may include one or more of the following components: a processing component 1202, a memory 1204, a power component 1206, a multimedia component 1208, an audio component 1210, an input/output (I/O) interface 1212, a sensor component 1214, and a communications component 1216.

The processing component 1202 generally controls overall operation of the electronic device 1200, such as operations associated with display, telephone calls, data communications, camera operations, and recording operations. The processing component 1202 may include one or more processors 1220 to execute instructions to perform all or part of the steps of the methods described above. Further, the processing component 1202 may include one or more modules that facilitate interactions between the processing component 1202 and other components. For example, the processing component 1202 may include a multimedia module to facilitate interaction between the multimedia component 1208 and the processing component 1202.

The memory 1204 is configured to store various types of data to support operations at the electronic device 1200. Examples of such data include instructions for any application or method operating on the electronic device 1200, contact data, phonebook data, messages, pictures, video, and so forth. The memory 1204 may be implemented by any type or combination of volatile or non-volatile memory devices, such as Static Random Access Memory (SRAM), electrically Erasable Programmable Read Only Memory (EEPROM), erasable Programmable Read Only Memory (EPROM), programmable Read Only Memory (PROM), read Only Memory (ROM), magnetic memory, flash memory, magnetic disk, optical disk, or graphene memory.

The power supply assembly 1206 provides power to the various components of the electronic device 1200. The power supply components 1206 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the electronic device 1200.

The multimedia component 1208 includes a screen between the electronic device 1200 and a user that provides an output interface. In some embodiments, the screen may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensor may sense not only the boundary of a touch or slide action, but also the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 1208 includes a front camera and/or a rear camera. When the electronic device 1200 is in an operational mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data. Each front and rear camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

The audio component 1210 is configured to output and/or input audio signals. For example, the audio component 1210 includes a Microphone (MIC) configured to receive external audio signals when the electronic device 1200 is in an operational mode, such as a call mode, a recording mode, and a speech recognition mode. The received audio signals may be further stored in the memory 1204 or transmitted via the communications component 1216. In some embodiments, the audio component 1210 further comprises a speaker for outputting audio signals.

The I/O interface 1212 provides an interface between the processing component 1202 and peripheral interface modules, which may be a keyboard, click wheel, buttons, etc. These buttons may include, but are not limited to: homepage button, volume button, start button, and lock button.

The sensor assembly 1214 includes one or more sensors for providing status assessment of various aspects of the electronic device 1200. For example, the sensor assembly 1214 may detect an on/off state of the electronic device 1200, a relative positioning of the components, such as a display and keypad of the electronic device 1200, the sensor assembly 1214 may also detect a change in position of the electronic device 1200 or a component of the electronic device 1200, the presence or absence of a user's contact with the electronic device 1200, an orientation or acceleration/deceleration of the device 1200, and a change in temperature of the electronic device 1200. The sensor assembly 1214 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 1214 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 1214 may also include an acceleration sensor, a gyroscopic sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

The communications component 1216 is configured to facilitate communication between the electronic device 1200 and other devices, either wired or wireless. The electronic device 1200 may access a wireless network based on a communication standard, such as WiFi, an operator network (e.g., 2G, 3G, 4G, or 5G), or a combination thereof. In one exemplary embodiment, the communication component 1216 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communications component 1216 further includes a Near Field Communication (NFC) module to facilitate short range communications. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, infrared data association (IrDA) technology, ultra Wideband (UWB) technology, bluetooth (BT) technology, and other technologies.

In an exemplary embodiment, the electronic device 1200 may be implemented by one or more Application Specific Integrated Circuits (ASICs), digital Signal Processors (DSPs), digital Signal Processing Devices (DSPDs), programmable Logic Devices (PLDs), field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic elements for executing the methods described above.

In an exemplary embodiment, a computer-readable storage medium is also provided, such as memory 1204, including instructions executable by processor 1220 of electronic device 1200 to perform the above-described method. For example, the computer readable storage medium may be ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

In an exemplary embodiment, a computer program product is also provided, comprising instructions executable by the processor 1220 of the electronic device 1200 to perform the above method.

Fig. 13 is a block diagram illustrating an electronic device 1300 for media information quality prediction, according to an example embodiment. For example, the electronic device 1300 may be a server. Referring to fig. 13, the electronic device 1300 includes a processing component 1320 that further includes one or more processors and memory resources represented by memory 1322 for storing instructions, such as application programs, that can be executed by the processing component 1320. The application programs stored in memory 1322 may include one or more modules each corresponding to a set of instructions. Further, the processing component 1320 is configured to execute instructions to perform the methods described above.

The electronic device 1300 may further include: the power component 1324 is configured to perform power management of the electronic device 1300, the wired or wireless network interface 1326 is configured to connect the electronic device 1300 to a network, and the input output (I/O) interface 1328. Electronic device 1300 may operate an operating system based on storage memory 1322, such as Window13 13erver,Mac O13 X,Unix,Linux,FreeB13D or the like.

In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory 1322, including instructions executable by a processor of the electronic device 1300 to perform the above-described method. The storage medium may be a computer readable storage medium, which may be, for example, ROM, random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.

In an exemplary embodiment, a computer program product is also provided, comprising instructions executable by a processor of the electronic device 1300 to perform the above-described method.

It should be noted that the descriptions of the foregoing apparatus, the electronic device, the computer readable storage medium, the computer program product, and the like according to the method embodiments may further include other implementations, and the specific implementation may refer to the descriptions of the related method embodiments and are not described herein in detail.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This disclosure is intended to cover any adaptations, uses, or adaptations of the disclosure following the general principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for predicting quality of media information, comprising:

performing quality prediction according to the medium information to be predicted and the historical data information of the medium information to be predicted to obtain predicted delivery data of the medium information to be predicted, wherein the quality prediction is realized through the quality prediction network, and the quality prediction network comprises a feature extraction network and a prediction network;

The quality prediction is performed according to the media information to be predicted and the historical data information of the media information to be predicted, so as to obtain predicted delivery data of the media information to be predicted, including:

The material characteristics are predicted through the prediction network, and predicted delivery data of the medium information to be predicted are obtained; wherein the delivery data comprises data for characterizing the delivery quality of the media information.

2. The method of claim 1, wherein the historical data information comprises first delivery data of the media information to be predicted within a preset time range, and the feature extraction network comprises a first network and a second network;

3. The method of claim 2, wherein the historical data information further includes at least one associated media information corresponding to the same media information publisher as the media information to be predicted, and second delivery data for each of the associated media information within the preset time range;

4. The method of claim 3, wherein the feature extraction network further includes a fusion network, and the obtaining the material features of the media information to be predicted according to the first material features and the second material features includes:

5. The method of any one of claims 2 to 4, wherein the first network comprises at least one of a first feature extraction module, a second feature extraction module, a third feature extraction module, a fourth feature extraction module;

6. The method according to any one of claims 1 to 4, wherein before the quality prediction is performed based on the medium information to be predicted and the historical data information of the medium information to be predicted, the method further comprises:

training the quality prediction network based on the prediction loss.

7. A media information quality prediction apparatus, comprising:

The prediction unit is configured to perform quality prediction according to the media information to be predicted and the historical data information of the media information to be predicted to obtain predicted delivery data of the media information to be predicted, wherein the quality prediction is realized through the quality prediction network, and the quality prediction network comprises a feature extraction network and a prediction network;

The prediction unit includes:

The first prediction subunit is configured to execute prediction processing on the material characteristics through the prediction network to obtain predicted delivery data of the medium information to be predicted; wherein the delivery data comprises data for characterizing the delivery quality of the media information.

8. The apparatus according to claim 7, wherein the history data information includes first delivery data of the medium information to be predicted within a preset time range, the feature extraction network includes a first network and a second network, and the first feature extraction subunit includes:

9. The apparatus according to claim 8, wherein the history data information further includes at least one associated media information corresponding to the same media information publisher as the media information to be predicted, and second delivery data for each of the associated media information within the preset time range; the medium information quality prediction apparatus further includes:

the processing sub-module is further configured to perform:

10. The media information quality prediction apparatus of claim 9, wherein the feature extraction network further comprises a fusion network, the processing sub-module further configured to perform:

11. The media information quality prediction apparatus according to any one of claims 8 to 10, wherein the first network includes at least one of a first feature extraction module, a second feature extraction module, a third feature extraction module, and a fourth feature extraction module;

12. The medium information quality prediction apparatus according to any one of claims 7 to 10, characterized in that the medium information quality prediction apparatus further comprises:

The training unit is further configured to perform:

training the quality prediction network based on the prediction loss.

13. An electronic device, comprising:

A processor;

a memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the media information quality prediction method of any one of claims 1 to 6.

14. A computer readable storage medium, characterized in that instructions in the computer readable storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the medium information quality prediction method of any one of claims 1 to 6.

15. A computer program product comprising instructions which, when executed by a processor of an electronic device, enable the electronic device to perform the medium information quality prediction method of any one of claims 1 to 6.