Media asset value evaluation method, device and equipment

Abstract

The application provides a media asset value evaluation method, device and equipment. The method comprises the following steps: obtaining historical characteristic data of N media assets, wherein N is an integer greater than or equal to 1; performing time series prediction on each media asset according to the historical characteristic data to obtain a future value grade of each media asset; performing trend analysis on each media asset according to the historical characteristic data to obtain a historical trend of each media asset; and obtaining a value classification label of each media asset according to the future value grade and the historical trend. The application considers the historical and future trend performance of media asset content in the whole life cycle, measures the value of the media asset by using the historical trend and the future value, solves the problem that the inaccurate value judgment of the media asset is caused by the historical performance considered by manual arrangement, can predict the recall set of intelligent recommendation, and is beneficial to exposing high-quality and high-value media assets to users and improving the click and conversion effect of the media asset.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and in particular to a method, apparatus and equipment for evaluating the value of media assets. Background Technology

[0002] Currently, as the main gateway to home entertainment, smart TVs rely on the media content and interactive entertainment methods they provide. Whether the media content exposed to users meets their needs and interests is a key task for smart TV content operators and personalized recommendation algorithms. In this regard, it is crucial to provide business operators with good media content arrangement or to provide recommendation algorithms with a high-quality recall set.

[0003] Existing media asset screening and recommendation technologies primarily utilize historical playback information to arrange and expose media assets. For example, Method 1, manual media asset screening, often relies on experience, limiting selection to a specific range based on viewing popularity and historical rankings. This narrow selection range leads to inaccurate recommendations. Method 2, personalized media asset recommendations relying on intelligent recommendation capabilities, typically requires a large database of media viewed and played historically. This database generates a subset of media assets matching user preferences, potentially resulting in a large recall set but a limited pool of high-quality content, thus impacting recommended content. Summary of the Invention

[0004] The purpose of this invention is to provide a method, apparatus, and equipment for evaluating the value of media assets, which solves the problem of inaccurate recommendation results in existing media asset screening schemes.

[0005] To achieve the above objectives, embodiments of the present invention provide a media asset value evaluation method, comprising:

[0006] Obtain historical feature data for N media assets, where N is an integer greater than or equal to 1;

[0007] Based on the historical feature data, time-series prediction is performed on each media asset to obtain the future value level of each media asset;

[0008] Based on the historical feature data, a trend analysis is performed on each media asset to obtain the historical trend of each media asset.

[0009] Based on the stated future value level and the stated historical trend, a value grading label is obtained for each of the media assets.

[0010] Optionally, the historical feature data includes: collection data, search data, playback data, and order data.

[0011] Optionally, the step of performing time-series prediction on each media asset based on the historical feature data to obtain the future value level of each media asset includes:

[0012] Test data is constructed based on the historical feature data of each media asset, and the test data includes feature data of M historical periods for each media asset;

[0013] The test data is input into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted.

[0014] Based on the predicted values ​​of the feature data, the media assets are classified into different levels to obtain the future value level of each media asset.

[0015] Optionally, constructing test data based on historical feature data of each media asset includes:

[0016] Based on the historical feature data of each media asset, a predictive feature matrix is ​​constructed for each media asset.

[0017] The predicted feature matrices corresponding to the N media assets are concatenated to obtain the feature matrix of the test data.

[0018] Optionally, the method further includes:

[0019] Training data is constructed based on the historical feature data of each media asset;

[0020] The prediction model is obtained by training the long short-term LSTM model using the training data.

[0021] Optionally, constructing training data based on the historical feature data of each media asset includes:

[0022] The feature data of each media asset for L historical periods are transformed to obtain a first matrix with supervised learning.

[0023] The first matrices corresponding to the N media assets are concatenated to obtain the feature matrix of the training data.

[0024] The first target data in the last column of the feature matrix is ​​deleted to obtain the training data.

[0025] Optionally, inputting the test data into the prediction model to obtain the predicted value of the feature data for the (M+1)th period includes:

[0026] The test data is input into the prediction model to obtain a first predicted value;

[0027] The first predicted value is inverted and scaled to obtain the predicted value of the feature data.

[0028] Optionally, the step of classifying the media assets based on the predicted values ​​of the feature data to obtain the future value level of each media asset includes:

[0029] Based on the relationship between the predicted values ​​of the feature data and the classification thresholds corresponding to different value levels, the future value levels of each media asset are classified.

[0030] Optionally, the future value level includes at least one of the following:

[0031] High-value media assets;

[0032] Popular media content;

[0033] Low-value media assets;

[0034] Media that attracts investment.

[0035] Optionally, trend analysis is performed on each media asset based on the historical feature data to obtain the historical trend of each media asset, including:

[0036] By using a predetermined verification method, the target historical trend of each media asset is obtained based on the second target data in the historical feature data;

[0037] The second target data includes playback data and / or order data. If the second target data is order data, then the target historical trend is the order trend; if the second target data is playback data, then the target historical trend is the playback trend.

[0038] Optionally, obtaining the target historical trend for each media asset based on the second target data in the historical feature data includes:

[0039] Calculate the test statistic based on the second target data;

[0040] Calculate the variance based on the test statistic;

[0041] Calculate the Z-statistic based on the test statistic and the variance;

[0042] The target historical trend of the media asset is determined based on the Z-statistic.

[0043] Optionally, the target historical trend includes one of the following:

[0044] Upward trend;

[0045] Downward trend;

[0046] Concave shape;

[0047] Convex shape;

[0048] No trend.

[0049] Optionally, the method further includes:

[0050] When the target historical trend is trendless, determine the concavity and convexity of the binomial fitting function corresponding to the media asset;

[0051] The target historical trend of the media asset is determined based on the concavity / convexity.

[0052] Optionally, obtaining historical feature data of N media assets includes:

[0053] Obtain initial data from different dimensions of N media assets;

[0054] The initial data is preprocessed to obtain the historical feature data;

[0055] The preprocessing includes missing value imputation and / or feature scaling.

[0056] To achieve the above objectives, embodiments of the present invention provide a media asset value evaluation device, comprising:

[0057] The first acquisition module is used to acquire historical feature data of N media assets, where N is an integer greater than or equal to 1;

[0058] The future value prediction module is used to perform time-series prediction on each media asset based on the historical feature data to obtain the future value level of each media asset.

[0059] The historical trend analysis module is used to perform trend analysis on each media asset based on the historical feature data, and to obtain the historical trend of each media asset.

[0060] The first determining module is used to determine the value grading label for each media asset based on the future value level and the historical trend.

[0061] Optionally, the historical feature data includes: collection data, search data, playback data, and order data.

[0062] Optionally, the future value prediction module includes:

[0063] The first construction unit is used to construct test data based on the historical feature data of each media asset, wherein the test data includes feature data of M historical periods for each media asset;

[0064] The prediction unit is used to input the test data into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted.

[0065] The grading unit is used to grade the media assets based on the predicted values ​​of the feature data, and to obtain the future value grade of each media asset.

[0066] Optionally, the first building unit includes:

[0067] The first conversion subunit is used to construct a prediction feature matrix for each media asset based on the historical feature data of each media asset.

[0068] The first splicing subunit is used to splice the prediction feature matrices corresponding to the N media assets to obtain the feature matrix of the test data.

[0069] Optionally, the device further includes:

[0070] The second construction unit is used to construct training data based on the historical feature data of each media asset;

[0071] The model training unit is used to train the long short-term LSTM model using the training data to obtain the prediction model.

[0072] Optionally, the second building unit includes:

[0073] The second transformation subunit is used to transform the feature data of L historical periods of each media asset to obtain a first matrix with supervised learning.

[0074] The second splicing subunit is used to splice the first matrices corresponding to the N media assets to obtain the feature matrix of the training data.

[0075] The deletion sub-unit is used to delete the first target data in the last column of the feature matrix to obtain the training data.

[0076] Optionally, the prediction unit includes:

[0077] A prediction subunit is used to input the test data into the prediction model to obtain a first prediction value;

[0078] The processing subunit is used to perform inversion scaling on the first predicted value to obtain the predicted value of the feature data.

[0079] Optionally, the grading unit is specifically used to: classify the future value level of each media asset based on the relationship between the predicted value of the feature data and the grading thresholds corresponding to different value levels.

[0080] Optionally, the future value level includes at least one of the following:

[0081] High-value media assets;

[0082] Popular media content;

[0083] Low-value media assets;

[0084] Media that attracts investment.

[0085] Optionally, the historical trend analysis module includes:

[0086] The first acquisition unit is used to acquire the target historical trend of each media asset based on the second target data in the historical feature data through a predetermined verification method.

[0087] The second target data includes playback data and / or order data. If the second target data is order data, then the target historical trend is the order trend; if the second target data is playback data, then the target historical trend is the playback trend.

[0088] Optionally, the first acquisition unit includes:

[0089] The first calculation subunit is used to calculate the test statistic based on the second target data;

[0090] The second calculation subunit is used to calculate the variance based on the test statistic.

[0091] The third calculation subunit is used to calculate the Z-statistic based on the test statistic and the variance.

[0092] A sub-unit is defined for determining the target historical trend of the media asset based on the Z statistic.

[0093] Optionally, the target historical trend includes one of the following:

[0094] Upward trend;

[0095] Downward trend;

[0096] Concave shape;

[0097] Convex shape;

[0098] No trend.

[0099] Optionally, the device further includes:

[0100] The second determining module is used to determine the concavity and convexity of the binomial fitting function corresponding to the media asset when the target historical trend is trendless.

[0101] The third determining module is used to determine the target historical trend of the media asset based on the concavity / convexity.

[0102] Optionally, the first acquisition module includes:

[0103] The second acquisition unit is used to acquire initial data from different dimensions of N media assets;

[0104] The preprocessing unit is used to preprocess the initial data to obtain the historical feature data;

[0105] The preprocessing includes missing value imputation and / or feature scaling.

[0106] To achieve the above objectives, embodiments of the present invention provide a media asset value evaluation device, comprising: a transceiver and a processor;

[0107] The transceiver is used to: acquire historical feature data of N media assets, where N is an integer greater than or equal to 1;

[0108] The processor is configured to: perform time-series prediction on each media asset based on the historical feature data, and obtain the future value level of each media asset;

[0109] Based on the historical feature data, a trend analysis is performed on each media asset to obtain the historical trend of each media asset.

[0110] Based on the stated future value level and the stated historical trend, a value grading label is obtained for each of the media assets.

[0111] Optionally, the historical feature data includes: collection data, search data, playback data, and order data.

[0112] Optionally, the processor performs time-series prediction on each media asset based on the historical feature data to obtain the future value level of each media asset, including:

[0113] Test data is constructed based on the historical feature data of each media asset, and the test data includes feature data of M historical periods for each media asset;

[0114] The test data is input into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted.

[0115] Based on the predicted values ​​of the feature data, the media assets are classified into different levels to obtain the future value level of each media asset.

[0116] Optionally, the processor constructs test data based on historical feature data of each media asset, including:

[0117] Based on the historical feature data of each media asset, a predictive feature matrix is ​​constructed for each media asset.

[0118] The predicted feature matrices corresponding to the N media assets are concatenated to obtain the feature matrix of the test data.

[0119] Optionally, the processor is further configured to:

[0120] Training data is constructed based on the historical feature data of each media asset;

[0121] The prediction model is obtained by training the long short-term LSTM model using the training data.

[0122] Optionally, the processor constructs training data based on historical feature data of each media asset, including:

[0123] The feature data of each media asset for L historical periods are transformed to obtain a first matrix with supervised learning.

[0124] The first matrices corresponding to the N media assets are concatenated to obtain the feature matrix of the training data.

[0125] The first target data in the last column of the feature matrix is ​​deleted to obtain the training data.

[0126] Optionally, the processor inputs the test data into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted, including:

[0127] The test data is input into the prediction model to obtain a first predicted value;

[0128] The first predicted value is inverted and scaled to obtain the predicted value of the feature data.

[0129] Optionally, the processor classifies the media assets into different levels based on the predicted values ​​of the feature data to obtain the future value level of each media asset, including:

[0130] Based on the relationship between the predicted values ​​of the feature data and the classification thresholds corresponding to different value levels, the future value levels of each media asset are classified.

[0131] Optionally, the future value level includes at least one of the following:

[0132] High-value media assets;

[0133] Popular media content;

[0134] Low-value media assets;

[0135] Media that attracts investment.

[0136] Optionally, the processor performs trend analysis on each media asset based on the historical feature data to obtain the historical trend of each media asset, including:

[0137] By using a predetermined verification method, the target historical trend of each media asset is obtained based on the second target data in the historical feature data;

[0138] The second target data includes playback data and / or order data. If the second target data is order data, then the target historical trend is the order trend; if the second target data is playback data, then the target historical trend is the playback trend.

[0139] Optionally, the processor obtains the target historical trend of each media asset based on the second target data in the historical feature data, including:

[0140] Calculate the test statistic based on the second target data;

[0141] Calculate the variance based on the test statistic;

[0142] Calculate the Z-statistic based on the test statistic and the variance;

[0143] The target historical trend of the media asset is determined based on the Z-statistic.

[0144] Optionally, the target historical trend includes one of the following:

[0145] Upward trend;

[0146] Downward trend;

[0147] Concave shape;

[0148] Convex shape;

[0149] No trend.

[0150] Optionally, the processor is further configured to: determine the concavity / convexity of the binomial fitting function corresponding to the media asset when the target historical trend is trendless;

[0151] The target historical trend of the media asset is determined based on the concavity / convexity.

[0152] Optionally, when the transceiver acquires historical feature data of N media assets, it is specifically used to: acquire initial data of different dimensions of the N media assets;

[0153] The processor is further configured to: preprocess the initial data to obtain the historical feature data;

[0154] The preprocessing includes missing value imputation and / or feature scaling.

[0155] To achieve the above objectives, embodiments of the present invention provide an electronic device, including: a transceiver, a processor, a memory, and a program or instructions stored in the memory and executable on the processor; characterized in that the processor implements the above-described media asset value evaluation method when executing the program or instructions.

[0156] To achieve the above objectives, embodiments of the present invention provide a readable storage medium having a program or instructions stored thereon, which, when executed by a processor, implement the steps of the media asset value evaluation method described above.

[0157] The beneficial effects of the above-described technical solution of the present invention are as follows:

[0158] The embodiments of the present invention integrate the historical and future trend performance of media asset content at a certain point in time throughout its entire life cycle, and use the dual factors of historical trend and future value to measure the value of media assets. This not only solves the problem that manual arrangement only considers historical performance, which leads to inaccurate value judgment of media assets, but also allows for the prediction of the recall set of intelligent recommendation, which is conducive to exposing high-quality and high-value media assets to users and improving the click-through and conversion effects of media assets. Attached Figure Description

[0159] Figure 1 This is one of the flowcharts illustrating the media asset value evaluation method according to an embodiment of the present invention;

[0160] Figure 2 This is a schematic diagram illustrating the classification of the future value of media assets according to an embodiment of the present invention;

[0161] Figure 3 This is a second schematic flowchart of the media asset value evaluation method according to an embodiment of the present invention;

[0162] Figure 4 This is a schematic diagram of the feature matrix according to an embodiment of the present invention;

[0163] Figure 5 This is a schematic diagram of the structure of the media asset value evaluation device according to an embodiment of the present invention;

[0164] Figure 6 This is a schematic diagram of the structure of the media asset value evaluation device according to an embodiment of the present invention;

[0165] Figure 7 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0166] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0167] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the invention. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0168] In various embodiments of the present invention, it should be understood that the sequence number of each process described below does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0169] In addition, the terms "system" and "network" are often used interchangeably in this article.

[0170] In the embodiments provided in this application, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean determining B solely based on A; B can also be determined based on A and / or other information.

[0171] like Figure 1 As shown, this embodiment of the invention provides a method for evaluating the value of media assets, including:

[0172] Step 11: Obtain historical feature data for N media assets, where N is an integer greater than or equal to 1.

[0173] The historical feature data may include: collection data, search data, playback data, and subscription data. The historical feature data is exemplified by: O = [F, S, P, D], where F = (f... ij ) N×T Represents the collection data, S = (s ij ) N×T Represents the search data, P = (p ij ) N×T Represents playback data, D = (d ij ) N×T This represents the order data; N represents the quantity of media assets, and T represents the number of historical periods for each media asset. i O represents the original feature matrix of the i-th media asset. i =[f i s i p i d i ], i = 1, 2…N; O it O represents the original feature matrix of the i-th media asset at historical moment j. ij =[f ij s ijp ij d ij ], j = 1, 2...T.

[0174] Step 12: Perform time-series prediction on each media asset based on the historical feature data to obtain the future value level of each media asset.

[0175] The future value rating of the media asset can be determined by using historical feature data to predict the feature data of the media asset for a future period (e.g., predicting the feature data of the (M+1)th period using data from M historical periods). The historical period could be, for example, the weekly peak playback time of a media asset (reaching a peak playback time every 7 days).

[0176] Step 13: Perform trend analysis on each media asset based on the historical feature data to obtain the historical trend of each media asset.

[0177] Trend analysis of the media asset can be performed to analyze the trend of a certain characteristic data of the media asset. For example, by using the historical playback data and subscription data of the media asset, the playback trend and subscription trend of the media asset can be analyzed, thereby determining the historical trend type of the media asset.

[0178] It should be noted that the execution order of steps 12 and 13 is not limited; they can be executed simultaneously, or steps 12 can be executed first and then steps 13, or vice versa. Figure 1 For illustrative purposes only.

[0179] Step 14: Obtain the value grading label for each media asset based on the future value level and the historical trend.

[0180] In this embodiment, based on the future value level and historical trend of the media assets, and taking into account the analysis results of the media assets' history and future prospects, the output media asset value classification tags can assist operators in arranging content. For example, for media assets predicted to be of high value, if the historical trend is also upward, the media asset may contribute to increased revenue or viewership. When arranging corresponding programs, the media asset can be placed in a more prominent position. If a media asset is predicted to be of low value and the historical trend is also downward, the media asset can be taken offline to avoid wasting recommended resources. Similarly, according to different needs, media assets are arranged according to their value classification tags.

[0181] The embodiments of the present invention integrate the historical and future trend performance of media asset content at a certain point in time throughout its entire life cycle, and use the dual factors of historical trend and future value to measure the value of media assets. This not only solves the problem that manual arrangement only considers historical performance, which leads to inaccurate value judgment of media assets, but also allows for the prediction of the recall set of intelligent recommendation, which is conducive to exposing high-quality and high-value media assets to users and improving the click-through and conversion effects of media assets.

[0182] Optionally, obtaining historical feature data of N media assets includes:

[0183] Obtain initial data from N media assets across different dimensions; preprocess the initial data to obtain the historical feature data; wherein the preprocessing includes missing value imputation and / or feature scaling.

[0184] In this embodiment, the initial data of different dimensions may include different types of data such as media asset collection, search, playback, and subscription. Specifically, the preprocessing of the initial data may include: imputing missing values ​​and / or scaling features for different dimensions of each media asset.

[0185] The missing value imputation refers to the process where, due to issues with data acquisition channels or transmission, some media assets may have missing data for a particular day. In such cases, the missing values ​​for that day can be filled using data from previous dates, for example, by selecting the average of the last seven periods prior to that date. It should be noted that during data preprocessing, this imputation operation needs to be performed separately for each media asset across different dimensions.

[0186] Feature scaling refers to transforming the data, such as converting it into input data suitable for LSTM models, for example, by performing current function normalization (i.e., min-max normalization).

[0187]

[0188] Where x′ represents the initial data, and x represents the data obtained after normalization. Therefore, the initial data of the media asset is, for example, O′=[F′,S′,P′,D′]. After data preprocessing, the historical feature matrix after feature processing can be obtained as: O=[F,S,P,D].

[0189] As an optional embodiment, step 12 includes:

[0190] Step 121: Construct test data based on the historical feature data of each media asset, wherein the test data includes feature data of M historical periods for each media asset.

[0191] The test data is the input data for prediction, which is the historical data of the media asset. If it is necessary to predict the feature data of the (M+1)th period, the test data is the historical feature data of the M periods before the (M+1)th period, where M is less than or equal to T.

[0192] Optionally, constructing test data based on the historical feature data of each media asset includes: constructing a prediction feature matrix for each media asset based on the historical feature data of each media asset; and concatenating the prediction feature matrices corresponding to the N media assets to obtain the feature matrix of the test data.

[0193] In this embodiment, the prediction feature matrix for media asset i can be represented as:

[0194]

[0195] For each of the N media assets that needs to be predicted, the above-mentioned prediction feature matrix is ​​calculated, and the N prediction feature matrices are concatenated row by row to obtain the feature matrix of the test data, which is used as the input of the prediction model. By inputting the trained prediction model, the predicted feature data value of each media asset in the M+1th period can be obtained.

[0196] Step 122: Input the test data into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted;

[0197] Optionally, the step of inputting the test data into the prediction model to obtain the predicted value of the feature data for the (M+1)th period includes: inputting the test data into the prediction model to obtain a first predicted value; and performing inversion scaling on the first predicted value to obtain the predicted value of the feature data.

[0198] In this embodiment, the historical feature data of the N media assets obtained can be preprocessed feature data, such as feature scaling. Therefore, when performing feature data prediction, the predicted value obtained after inputting the test data into the prediction model needs to undergo inverse scaling processing to obtain the final feature data prediction value.

[0199] The prediction model can be obtained through model training. Optionally, the method further includes: constructing training data based on the historical feature data of each media asset; and training a long short-term LSTM model using the training data to obtain the prediction model.

[0200] The step of constructing training data based on the historical feature data of each media asset may include: performing data transformation on the feature data of L historical periods of each media asset to obtain a first matrix with supervised learning; concatenating the first matrices corresponding to N media assets to obtain the feature matrix of the training data; and deleting the first target data in the last column of the feature matrix to obtain the training data.

[0201] The first target data can be data that does not need to be predicted. For example, in an embodiment of the present invention, only the number of plays and subscription revenue are predicted. In this case, the first target data is collection feature data and search feature data. After deleting the collection feature data and search feature data in the last column of the matrix, only the columns that need to be predicted are retained.

[0202] Suppose we use historical feature data from L out of T historical periods for model training, where the historical feature data from the Lth period is used as the true value for the period to be predicted, and the historical feature data from the remaining L-1 periods are used as the input values ​​for predicting the feature data of the Lth period. Then, for each media resource i, the feature matrix O i The first matrix obtained after the transformation can be in the following form:

[0203]

[0204] The above transformation is performed on all media assets, and they are then concatenated row by row to obtain the feature matrix X = (O ij ) M×K Where M = N*(TL), K = (L+1), the feature matrix of the training data can also be expressed in the following form:

[0205] X = [x L x L-1 [x1, x0]

[0206] Each x can include four columns of data: favorites, searches, plays, and subscriptions.

[0207] After obtaining the above feature matrix, the LSTM model is trained using the training data. In an embodiment of the present invention, only the number of plays and subscription revenue can be predicted. In this case, the last group x0 of the above feature matrix X needs to be processed, and only the columns that need to be predicted are retained. That is, the collection feature and search feature data in x0 are deleted. The processed training data is then input into the LSTM model for model training.

[0208] Step 123: Based on the predicted values ​​of the feature data, classify the media assets into different levels to obtain the future value level of each media asset.

[0209] Specifically, the step of classifying the media assets based on the predicted values ​​of the feature data to obtain the future value level of each media asset may include: classifying the future value level of each media asset according to the relationship between the predicted values ​​of the feature data and the classification thresholds corresponding to different value levels. It should be noted that when classifying the future value level, the classification can be based on the predicted values ​​of each media asset's playback data and subscription data.

[0210] The future value level may include at least one of the following:

[0211] High-value media assets;

[0212] Popular media content;

[0213] Low-value media assets;

[0214] Media that attracts investment.

[0215] In this embodiment, thresholds can be set for different value levels. For example, by dividing thresholds, the future value levels can be divided as follows: Figure 2 The four categories shown are as follows: those with more than the first value and more than the second value are high-value media assets; those with more than the first value but less than the second value are popular media assets; those with less than the first value and less than the second value are low-value media assets; and those with less than the first value but more than the second value are money-making media assets.

[0216] As an optional embodiment, step 13 may include: obtaining the target historical trend of each media asset based on the second target data in the historical feature data using a predetermined verification method;

[0217] The second target data includes playback data and / or order data. If the second target data is order data, then the target historical trend is the order trend; if the second target data is playback data, then the target historical trend is the playback trend.

[0218] This embodiment analyzes the historical trends of each media asset based on the historical feature data obtained in step 11. During the historical trend analysis, secondary target data for each media asset can be analyzed, such as playback data and / or subscription data, to obtain the playback trend and / or subscription trend of the media asset. The predetermined verification method is, for example, the MK test method, which determines the target historical trend of each media asset based on the MK test.

[0219] Optionally, obtaining the target historical trend for each media asset based on the second target data in the historical feature data includes:

[0220] Step 131: Calculate the test statistic based on the second target data.

[0221] Taking the second target data as playback data as an example, suppose the playback time series of a certain media asset for M historical periods is: (a1, a2, ..., a M For all k, y≤M and k≠y, k=1、…、M-1, y=k+1、…、M, calculate the test statistic S as follows:

[0222]

[0223] in,

[0224] Step 132: Calculate the variance based on the test statistic.

[0225] Specifically, the formula for calculating the variance V is as follows:

[0226] V = M(M-1)(2M+5) / 18

[0227] Step 133: Calculate the Z-statistic based on the test statistic and the variance.

[0228] Specifically, the formula for calculating the Z statistic is as follows:

[0229]

[0230] Step 134: Determine the target historical trend of the media asset based on the Z statistic.

[0231] Specifically, the target historical trend may include one of the following:

[0232] An upward trend;

[0233] Decreasing trend;

[0234] A concave shape, meaning it first descends and then rises;

[0235] A convex pattern, meaning it rises first and then falls;

[0236] No trend.

[0237] Furthermore, based on the Z-statistic, the target historical trend of the media asset is determined as follows:

[0238]

[0239] Optionally, the method further includes: when the target historical trend is trendless, determining the concavity / convexity of the binomial fitting function corresponding to the media asset; and determining the target historical trend of the media asset based on the concavity / convexity.

[0240] In this embodiment, when the target historical trend of the media asset is determined to be trendless based on the Z-statistic, the concavity / convexity of the trendless media asset can be further determined according to the binomial fitted function. The historical trend of the media asset is then determined based on the concavity / convexity of the function. It should be noted that when determining the historical trend of the media asset based on concavity / convexity, the obtained result may be one of three cases: concave, convex, or trendless. Thus, media assets can be classified into five trend types based on historical trend: rising, falling, concave, convex, and trendless.

[0241] In embodiments of the present invention, based on the prediction of the future status of media assets obtained in step 12 and the analysis of the historical trends of media assets obtained in step 13, the value classification labels of media assets are output by comprehensively considering the analysis results of the past and future to assist operators in arranging content. For example, for media assets with a high future value classification, if the historical trend also shows a double upward trend (both the trends of play count and subscription revenue are upward), then the media asset is more important for increasing revenue or viewership, and can be placed in a prominent position when arranging corresponding programs. If the media asset with a low future value classification also shows a double downward trend (both the trends of play count and subscription revenue are downward), then the media asset will not have a good effect on play count and revenue, and can be taken offline in a timely manner to avoid wasting recommended resources.

[0242] The implementation process of the media asset value evaluation method of this invention is illustrated below through specific embodiments. For example... Figure 3 As shown, it specifically includes five steps:

[0243] Step 31: Obtain data. This involves obtaining initial data from different dimensions of N media assets, which may include collection data, search data, playback data, and subscription data.

[0244] Step 32: Data Preprocessing. Missing values ​​are imputed and features are standardized on the initial data of each media asset to obtain historical feature data for each asset.

[0245] Step 33: Future Value Assessment. Multivariate prediction is performed using an LSTM model to predict the play count and subscription volume of N media assets. Based on the threshold values ​​corresponding to different value levels, each media asset is divided into multiple levels. For example, the N media assets can be divided into high, medium, and low value assets based on an RFM model.

[0246] Step 34: Historical Trend Analysis. For the preprocessed historical feature data, perform the MK test to determine whether there is a trend, and output the historical trend of each media asset. If it is determined that there is no trend, perform a concavity / convexity test on the media assets without a trend, and output the trend.

[0247] Step 35: Output the value grading labels for media assets. Based on the predicted future value and the analysis of historical trends, obtain the value grading labels for the media assets.

[0248] The following specific examples illustrate this:

[0249] Suppose there are three media assets, A, B, and C, and the value of each asset needs to be evaluated. The process for evaluating the value of these three media assets based on this application is as follows:

[0250] Step 1: Obtain the historical collection, search, playback, and subscription data O = [F, S, P, D] for each of the three media assets mentioned above, and complete the filling of missing values ​​and feature scaling.

[0251] Step 2: LSTM Prediction and Value Allocation:

[0252] Taking media asset A as an example, assuming the selected historical period number is 11, and the features of the previous 8 periods are selected as input to predict the situation of the next period (i.e., the 9th period):

[0253] (1) Data Construction: Preprocessed sales data such as Figure 4 As shown on the left, the four columns of data represent collection, search, playback, and order data, respectively. Using the features of the previous eight periods as input, a new feature matrix is ​​obtained to predict the next period, as shown below. Figure 4 As shown.

[0254] (2) LSTM time series prediction: All media asset data are decomposed into new data formats according to the data construction steps, and then all media assets are... Figure 4 The X and Y values ​​shown are concatenated to serve as the input and output of the model, predicting the viewership and subscription revenue of media assets A and B in the next period. It is assumed that the predicted viewership and subscription revenue values ​​of media asset A are [0.8 0.8], the predicted viewership and subscription revenue values ​​of media asset B are [0.3 0.2], and the predicted viewership and subscription revenue values ​​of media asset C are [0.6 0.4].

[0255] (3) Based on the set threshold, assuming that the thresholds for playback and subscription are both set to 0.5, then media asset A is a high-value media asset, media asset B is a low-value media asset, and media asset C is a popular media asset.

[0256] Step 3: Analyze historical trends based on media asset playback and subscription data. Let's take media asset A's playback and subscription data as an example for analysis:

[0257] MK test: The playback data of media asset A shows an upward trend, while the subscription data shows a concave shape. Therefore, media asset A is a type of media asset with a double increase (the number of playbacks increases, while subscription revenue decreases first and then increases).

[0258] Step 4: After completing the above three steps, you can obtain the specific value classification label of the media asset. For example, media asset A is a high-value media asset with both positive and negative aspects, and you can consider giving it more frequent exposure.

[0259] To illustrate with a comparative example: Actual media asset A is a periodically updated media asset, with a peak occurring every 7 days. Figure 4 (Only a portion of the data was extracted). Based on LSTM, this periodic pattern can be learned, so the predicted results in step (2) are relatively high: the predicted values ​​for viewership and subscription revenue are [0.8 0.8] respectively. However, if only the average level of the media asset's historical performance is used, the average values ​​for viewership and subscription revenue are only [0.3 0.3], which will underestimate the value of the media asset. Therefore, predicting the future situation of the media asset based on historical behavior can more accurately assess the future value of the media asset. In addition, historical trend information is also used to more accurately assess the media asset. For example, for daily updated TV series, after the update is completed, the viewership and subscription revenue of the media asset will still remain at a certain level, but compared with its own historical situation, it will begin to show a downward trend. By judging the trend, the scheduling of the media asset can be adjusted in a timely manner.

[0260] The embodiments of the present invention integrate the historical and future trend performance of media asset content at a certain point in time throughout its entire life cycle, and use the dual factors of historical trend and future value to measure the value of media assets. This not only solves the problem that manual arrangement only considers historical performance, which leads to inaccurate value judgment of media assets, but also allows for the prediction of the recall set of intelligent recommendation, which is conducive to exposing high-quality and high-value media assets to users and improving the click-through and conversion effects of media assets.

[0261] like Figure 5 As shown, this embodiment of the invention also provides a media asset value evaluation device 500, comprising:

[0262] The first acquisition module 510 is used to acquire historical feature data of N media assets, where N is an integer greater than or equal to 1;

[0263] The future value prediction module 520 is used to perform time-series prediction on each media asset based on the historical feature data to obtain the future value level of each media asset.

[0264] The historical trend analysis module 530 is used to perform trend analysis on each media asset based on the historical feature data, and to obtain the historical trend of each media asset.

[0265] The first determining module 540 is used to determine the value grading label of each media asset based on the future value level and the historical trend.

[0266] Optionally, the historical feature data includes: collection data, search data, playback data, and order data.

[0267] Optionally, the future value prediction module includes:

[0268] The first construction unit is used to construct test data based on the historical feature data of each media asset, wherein the test data includes feature data of M historical periods for each media asset;

[0269] The prediction unit is used to input the test data into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted.

[0270] The grading unit is used to grade the media assets based on the predicted values ​​of the feature data, and to obtain the future value grade of each media asset.

[0271] Optionally, the first building unit includes:

[0272] The first conversion subunit is used to construct a prediction feature matrix for each media asset based on the historical feature data of each media asset.

[0273] The first splicing subunit is used to splice the prediction feature matrices corresponding to the N media assets to obtain the feature matrix of the test data.

[0274] Optionally, the device further includes:

[0275] The second construction unit is used to construct training data based on the historical feature data of each media asset;

[0276] The model training unit is used to train the long short-term LSTM model using the training data to obtain the prediction model.

[0277] Optionally, the second building unit includes:

[0278] The second transformation subunit is used to transform the feature data of L historical periods of each media asset to obtain a first matrix with supervised learning.

[0279] The second splicing subunit is used to splice the first matrices corresponding to the N media assets to obtain the feature matrix of the training data.

[0280] The deletion sub-unit is used to delete the first target data in the last column of the feature matrix to obtain the training data.

[0281] Optionally, the prediction unit includes:

[0282] A prediction subunit is used to input the test data into the prediction model to obtain a first prediction value;

[0283] The processing subunit is used to perform inversion scaling on the first predicted value to obtain the predicted value of the feature data.

[0284] Optionally, the grading unit is specifically used to: classify the future value level of each media asset based on the relationship between the predicted value of the feature data and the grading thresholds corresponding to different value levels.

[0285] Optionally, the future value level includes at least one of the following:

[0286] High-value media assets;

[0287] Popular media content;

[0288] Low-value media assets;

[0289] Media that attracts investment.

[0290] Optionally, the historical trend analysis module includes:

[0291] The first acquisition unit is used to acquire the target historical trend of each media asset based on the second target data in the historical feature data through a predetermined verification method.

[0292] The second target data includes playback data and / or order data. If the second target data is order data, then the target historical trend is the order trend; if the second target data is playback data, then the target historical trend is the playback trend.

[0293] Optionally, the first acquisition unit includes:

[0294] The first calculation subunit is used to calculate the test statistic based on the second target data;

[0295] The second calculation subunit is used to calculate the variance based on the test statistic.

[0296] The third calculation subunit is used to calculate the Z-statistic based on the test statistic and the variance.

[0297] A sub-unit is defined for determining the target historical trend of the media asset based on the Z statistic.

[0298] Optionally, the target historical trend includes one of the following:

[0299] Upward trend;

[0300] Downward trend;

[0301] Concave shape;

[0302] Convex shape;

[0303] No trend.

[0304] Optionally, the device further includes:

[0305] The second determining module is used to determine the concavity and convexity of the binomial fitting function corresponding to the media asset when the target historical trend is trendless.

[0306] The third determining module is used to determine the target historical trend of the media asset based on the concavity / convexity.

[0307] Optionally, the first acquisition module includes:

[0308] The second acquisition unit is used to acquire initial data from different dimensions of N media assets;

[0309] The preprocessing unit is used to preprocess the initial data to obtain the historical feature data;

[0310] The preprocessing includes missing value imputation and / or feature scaling.

[0311] It should be noted that the media asset value evaluation device provided in this embodiment of the invention can implement all the method steps implemented in the above-mentioned media asset value evaluation method embodiment, and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiment and the beneficial effects will not be described in detail.

[0312] like Figure 6 As shown, an embodiment of the media asset value evaluation device 600 of the present invention includes a processor 610 and a transceiver 620, wherein,

[0313] The transceiver 620 is used to: acquire historical feature data of N media assets, where N is an integer greater than or equal to 1;

[0314] The processor 610 is configured to: perform time-series prediction on each media asset based on the historical feature data, and obtain the future value level of each media asset;

[0315] Based on the historical feature data, a trend analysis is performed on each media asset to obtain the historical trend of each media asset.

[0316] Based on the stated future value level and the stated historical trend, a value grading label is obtained for each of the media assets.

[0317] Optionally, the historical feature data includes: collection data, search data, playback data, and order data.

[0318] Optionally, the processor performs time-series prediction on each media asset based on the historical feature data to obtain the future value level of each media asset, including:

[0319] Test data is constructed based on the historical feature data of each media asset, and the test data includes feature data of M historical periods for each media asset;

[0320] The test data is input into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted.

[0321] Based on the predicted values ​​of the feature data, the media assets are classified into different levels to obtain the future value level of each media asset.

[0322] Optionally, the processor constructs test data based on historical feature data of each media asset, including:

[0323] Based on the historical feature data of each media asset, a predictive feature matrix is ​​constructed for each media asset.

[0324] The predicted feature matrices corresponding to the N media assets are concatenated to obtain the feature matrix of the test data.

[0325] Optionally, the processor is further configured to:

[0326] Training data is constructed based on the historical feature data of each media asset;

[0327] The prediction model is obtained by training the long short-term LSTM model using the training data.

[0328] Optionally, the processor constructs training data based on historical feature data of each media asset, including:

[0329] The feature data of each media asset for L historical periods are transformed to obtain a first matrix with supervised learning.

[0330] The first matrices corresponding to the N media assets are concatenated to obtain the feature matrix of the training data.

[0331] The first target data in the last column of the feature matrix is ​​deleted to obtain the training data.

[0332] Optionally, the processor inputs the test data into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted, including:

[0333] The test data is input into the prediction model to obtain a first predicted value;

[0334] The first predicted value is inverted and scaled to obtain the predicted value of the feature data.

[0335] Optionally, the processor classifies the media assets into different levels based on the predicted values ​​of the feature data to obtain the future value level of each media asset, including:

[0336] Based on the relationship between the predicted values ​​of the feature data and the classification thresholds corresponding to different value levels, the future value levels of each media asset are classified.

[0337] Optionally, the future value level includes at least one of the following:

[0338] High-value media assets;

[0339] Popular media content;

[0340] Low-value media assets;

[0341] Media that attracts investment.

[0342] Optionally, the processor performs trend analysis on each media asset based on the historical feature data to obtain the historical trend of each media asset, including:

[0343] By using a predetermined verification method, the target historical trend of each media asset is obtained based on the second target data in the historical feature data;

[0344] The second target data includes playback data and / or order data. If the second target data is order data, then the target historical trend is the order trend; if the second target data is playback data, then the target historical trend is the playback trend.

[0345] Optionally, the processor obtains the target historical trend of each media asset based on the second target data in the historical feature data, including:

[0346] Calculate the test statistic based on the second target data;

[0347] Calculate the variance based on the test statistic;

[0348] Calculate the Z-statistic based on the test statistic and the variance;

[0349] The target historical trend of the media asset is determined based on the Z-statistic.

[0350] Optionally, the target historical trend includes one of the following:

[0351] Upward trend;

[0352] Downward trend;

[0353] Concave shape;

[0354] Convex shape;

[0355] No trend.

[0356] Optionally, the processor is further configured to: determine the concavity / convexity of the binomial fitting function corresponding to the media asset when the target historical trend is trendless;

[0357] The target historical trend of the media asset is determined based on the concavity / convexity.

[0358] Optionally, when the transceiver acquires historical feature data of N media assets, it is specifically used to: acquire initial data of different dimensions of the N media assets;

[0359] The processor is further configured to: preprocess the initial data to obtain the historical feature data;

[0360] The preprocessing includes missing value imputation and / or feature scaling.

[0361] It should be noted that the media asset value evaluation device provided in the embodiments of the present invention can realize all the method steps implemented in the above media asset value evaluation method embodiments and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiments and the beneficial effects will not be described in detail.

[0362] Another embodiment of the electronic device of the present invention, such as Figure 7 As shown, it includes a transceiver 710, a processor 700, a memory 720, and a program or instructions stored in the memory 720 and executable on the processor 700; when the processor 700 executes the program or instructions, it implements the above-mentioned media asset value evaluation method.

[0363] The transceiver 710 is used to receive and send data under the control of the processor 700.

[0364] Among them, Figure 7In this context, the bus architecture may include any number of interconnected buses and bridges, specifically linking various circuits together, represented by one or more processors (processor 700) and memory (memory 720). The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides an interface. The transceiver 710 may be multiple elements, including transmitters and receivers, providing a unit for communicating with various other devices over a transmission medium. The processor 700 is responsible for managing the bus architecture and general processing, and the memory 720 may store data used by the processor 700 during operation.

[0365] An embodiment of the present invention provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the steps in the media asset value evaluation method described above and achieve the same technical effect. To avoid repetition, further details are omitted here.

[0366] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0367] It should be further noted that the electronic devices described in this specification include, but are not limited to, smartphones, tablets, etc., and many of the described functional components are referred to as modules in order to more specifically emphasize the independence of their implementation.

[0368] In this embodiment of the invention, the module can be implemented in software so that it can be executed by various types of processors. For example, an identified executable code module may include one or more physical or logical blocks of computer instructions, which may be constructed as objects, procedures, or functions. Nevertheless, the executable code of the identified module does not need to be physically located together, but may include different instructions stored in different bits, which, when logically combined, constitute the module and achieve the module's intended purpose.

[0369] In practice, an executable code module can be a single instruction or many instructions, and can even be distributed across multiple different code segments, different programs, and across multiple memory devices. Similarly, operational data can be identified within the module and can be implemented in any suitable form and organized within any suitable type of data structure. This operational data can be collected as a single dataset or distributed across different locations (including different storage devices), and can exist, at least in part, solely as electronic signals within the system or network.

[0370] When a module can be implemented using software, considering the current level of hardware technology, modules that can be implemented in software can be implemented using hardware circuits by those skilled in the art to achieve the corresponding functions, without considering cost. These hardware circuits include conventional very-large-scale integrated circuits (VLSI) or gate arrays, as well as existing semiconductors such as logic chips and transistors, or other discrete components. Modules can also be implemented using programmable hardware devices, such as field-programmable gate arrays, programmable array logic, and programmable logic devices.

[0371] The exemplary embodiments described above are with reference to the accompanying drawings. Many different forms and embodiments are feasible without departing from the spirit and teachings of the invention. Therefore, the invention should not be construed as limiting the exemplary embodiments set forth herein. Rather, these exemplary embodiments are provided to make the invention complete and convey the scope of the invention to those skilled in the art. In these drawings, component dimensions and relative dimensions may be exaggerated for clarity. The terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. As used herein, unless clearly indicated otherwise, the singular forms “a,” “an,” and “the” are intended to include all such forms. It will be further understood that the terms “comprising” and / or “including”, when used in this specification, indicate the presence of the stated features, integers, steps, operations, components, and / or elements, but do not exclude the presence or addition of one or more other features, integers, steps, operations, components, and / or groups thereof. Unless otherwise indicated, when stated, a range of values ​​includes the upper and lower limits of the range and any subranges in between.

[0372] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for evaluating the value of media assets, characterized in that, include: Obtain historical feature data for N media assets, where N is an integer greater than or equal to 1; the historical feature data includes: collection data, search data, playback data, and subscription data; Based on the historical feature data, time-series prediction is performed on each media asset to obtain the future value level of each media asset; Based on the historical feature data, a trend analysis is performed on each media asset to obtain the historical trend of each media asset. Based on the future value level and the historical trend, a value grading label is obtained for each media asset; The value rating tags for media assets are used to assist operators in arranging media assets. The value rating tags for media assets integrate the historical and future trend performance of media asset content at a certain point in time throughout its entire life cycle. The step of performing time-series prediction on each media asset based on the historical feature data to obtain the future value level of each media asset includes: constructing test data based on the historical feature data of each media asset, the test data including feature data of M historical periods for each media asset; inputting the test data into a prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted; classifying the media assets based on the predicted value of the feature data to obtain the future value level of each media asset; the step of inputting the test data into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted includes: inputting the test data into the prediction model to obtain a first predicted value; and performing inversion scaling on the first predicted value to obtain the predicted value of the feature data. Performing trend analysis on each media asset based on the historical feature data to obtain the historical trend of each media asset includes: obtaining the target historical trend of each media asset based on the second target data in the historical feature data using a predetermined testing method; wherein the second target data includes playback data and / or subscription data, if the second target data is subscription data, then the target historical trend is a subscription trend; if the second target data is playback data, then the target historical trend is a playback trend; obtaining the target historical trend of each media asset based on the second target data in the historical feature data includes: calculating a test statistic based on the second target data; calculating the variance based on the test statistic; calculating a Z-statistic based on the test statistic and the variance; and determining the target historical trend of the media asset based on the Z-statistic.

2. The method according to claim 1, characterized in that, The step of constructing test data based on the historical characteristic data of each media asset includes: Based on the historical feature data of each media asset, a predictive feature matrix is ​​constructed for each media asset. The predicted feature matrices corresponding to the N media assets are concatenated to obtain the feature matrix of the test data.

3. The method according to claim 1, characterized in that, The method further includes: Training data is constructed based on the historical feature data of each media asset; The prediction model is obtained by training the long short-term LSTM model using the training data.

4. The method according to claim 3, characterized in that, The step of constructing training data based on the historical feature data of each media asset includes: The feature data of each media asset for L historical periods are transformed to obtain a first matrix with supervised learning. The first matrices corresponding to the N media assets are concatenated to obtain the feature matrix of the training data. The first target data in the last column of the feature matrix is ​​deleted to obtain the training data.

5. The method according to claim 1, characterized in that, The step of classifying the media assets based on the predicted values ​​of the feature data to obtain the future value level of each media asset includes: Based on the relationship between the predicted values ​​of the feature data and the classification thresholds corresponding to different value levels, the future value levels of each media asset are classified.

6. The method according to claim 1, characterized in that, The future value level includes at least one of the following: High-value media assets; Popular media content; Low-value media assets; Media that attracts investment.

7. The method according to claim 1, characterized in that, The target historical trend includes one of the following: Upward trend; Downward trend; Concave shape; Convex shape; No trend.

8. The method according to claim 1, characterized in that, The method further includes: When the target historical trend is trendless, determine the concavity and convexity of the binomial fitting function corresponding to the media asset; The target historical trend of the media asset is determined based on the concavity / convexity.

9. The method according to claim 1, characterized in that, The acquisition of historical feature data for N media assets includes: Obtain initial data from different dimensions of N media assets; The initial data is preprocessed to obtain the historical feature data; The preprocessing includes missing value imputation and / or feature scaling.

10. A media asset value evaluation device, characterized in that, include: The first acquisition module is used to acquire historical feature data of N media assets, where N is an integer greater than or equal to 1; the historical feature data includes: collection data, search data, playback data, and subscription data; The future value prediction module is used to perform time-series prediction on each media asset based on the historical feature data to obtain the future value level of each media asset. The historical trend analysis module is used to perform trend analysis on each media asset based on the historical feature data, and to obtain the historical trend of each media asset. The determination module is used to determine the value grading label for each of the media assets based on the future value level and the historical trend; The value rating tags for media assets are used to assist operators in arranging media assets. The value rating tags for media assets integrate the historical and future trend performance of media asset content at a certain point in time throughout its entire life cycle. The future value prediction module includes: The first construction unit is used to construct test data based on the historical feature data of each media asset, wherein the test data includes feature data of M historical periods for each media asset; A prediction unit is used to input the test data into a prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted; the prediction unit includes: a prediction subunit, used to input the test data into the prediction model to obtain a first predicted value; and a processing subunit, used to perform inversion scaling on the first predicted value to obtain the predicted value of the feature data. A grading unit is used to grade the media assets based on the predicted values ​​of the feature data, thereby obtaining the future value grade of each media asset. The historical trend analysis module includes: a first acquisition unit, used to acquire the target historical trend of each media asset based on the second target data in the historical feature data using a predetermined testing method; wherein the second target data includes playback data and / or subscription data, and if the second target data is subscription data, the target historical trend is a subscription trend; if the second target data is playback data, the target historical trend is a playback trend; the first acquisition unit includes: a first calculation subunit, used to calculate a test statistic based on the second target data; a second calculation subunit, used to calculate the variance based on the test statistic; a third calculation subunit, used to calculate a Z-statistic based on the test statistic and the variance; and a determination subunit, used to determine the target historical trend of the media asset based on the Z-statistic.

11. A media asset value evaluation device, characterized in that, include: Transceiver and processor; The transceiver is used to: acquire historical feature data of N media assets, where N is an integer greater than or equal to 1; the historical feature data includes: collection data, search data, playback data, and subscription data; The processor is configured to: perform time-series prediction on each media asset based on the historical feature data, and obtain the future value level of each media asset; Based on the historical feature data, a trend analysis is performed on each media asset to obtain the historical trend of each media asset. Based on the future value level and the historical trend, a value grading label is obtained for each media asset; The value rating tags for media assets are used to assist operators in arranging media assets. The value rating tags for media assets integrate the historical and future trend performance of media asset content at a certain point in time throughout its entire life cycle. The processor performs time-series prediction on each media asset based on the historical feature data to obtain the future value level of each media asset, including: constructing test data based on the historical feature data of each media asset, the test data including feature data of M historical periods for each media asset; inputting the test data into a prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted; classifying the media assets based on the predicted value of the feature data to obtain the future value level of each media asset; the processor inputs the test data into the prediction model to obtain the predicted value of the feature data for the (M+1)th period to be predicted, including: inputting the test data into the prediction model to obtain a first predicted value; performing inversion scaling on the first predicted value to obtain the predicted value of the feature data. The processor performs trend analysis on each media asset based on the historical feature data to obtain the historical trend of each media asset, including: obtaining the target historical trend of each media asset based on the second target data in the historical feature data using a predetermined testing method; wherein the second target data includes playback data and / or subscription data, if the second target data is subscription data, then the target historical trend is a subscription trend; if the second target data is playback data, then the target historical trend is a playback trend; the processor obtains the target historical trend of each media asset based on the second target data in the historical feature data, including: calculating a test statistic based on the second target data; calculating the variance based on the test statistic; calculating a Z-statistic based on the test statistic and the variance; and determining the target historical trend of the media asset based on the Z-statistic.

12. An electronic device, comprising: A transceiver, a processor, a memory, and a program or instructions stored in the memory and executable on the processor; characterized in that, when the processor executes the program or instructions, it implements the media asset valuation method as described in any one of claims 1-9.

13. A readable storage medium having a program or instructions stored thereon, characterized in that, When the program or instructions are executed by the processor, they implement the steps of the media asset value evaluation method as described in any one of claims 1-9.

Images