User-level video data index acquisition method and device and medium

Abstract

The invention provides a user-level video data index acquisition method and device and a medium. The method comprises the following steps: acquiring video service deep packet inspection (DPI) data of a user; and inputting the video service DPI data of the user into N trained index prediction models, and predicting N user-level video data indexes of the user through the N index prediction models, the trained N index prediction models are obtained by training on the basis of historical video service DPI data of multiple users and N historical user-level video data indexes, and N is greater than or equal to 1. According to the method, the device and the medium, the problem that an existing user-level video data index acquisition method is limited by a base station manufacturer, so that index acquisition is difficult can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data acquisition, and in particular to a method, device and medium for acquiring user-level video data indicators. Background Art

[0002] With the widespread popularity of 5G communications, video data has become more and more common, and the amount of data has increased exponentially. How to obtain multiple user-level video data indicators of users through technical means, and then analyze the user's overall perception and quality has become an inevitable requirement for technological development.

[0003] However, the existing methods for obtaining user-level video data indicators are limited by base station manufacturers, making it difficult to obtain the indicators. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method, device and medium for obtaining user-level video data indicators in view of the above-mentioned deficiencies in the prior art, so as to solve the problem that the existing method for obtaining user-level video data indicators is limited by base station manufacturers, resulting in difficulty in obtaining indicators.

[0005] In a first aspect, the present invention provides a method for obtaining a user-level video data indicator, comprising:

[0006] Obtain the user's video service deep packet inspection DPI data;

[0007] The video service DPI data of the user is input into N trained indicator prediction models, and N user-level video data indicators of the user are predicted by the N indicator prediction models, wherein the trained N indicator prediction models are trained based on historical video service DPI data of multiple users and N historical user-level video data indicators, and N is greater than or equal to 1.

[0008] Further, the obtaining of the user's video service deep packet inspection DPI data specifically includes:

[0009] Obtaining the video service DPI data from the transport layer and the network layer of the network model;

[0010] The step of inputting the video service DPI data of the user into the trained N indicator prediction models, and predicting the N user-level video data indicators of the user by using the N indicator prediction models, specifically includes:

[0011] Extracting target DPI data having a high correlation with each of the user-level video data indicators from the acquired video service DPI data;

[0012] The target DPI data with a high correlation with each of the user-level video data indicators is respectively input into the corresponding trained indicator prediction model to obtain the predicted value of each of the user-level video data indicators.

[0013] Furthermore, before inputting the target DPI data having a high correlation with each of the user-level video data indicators into the corresponding trained indicator prediction model to obtain the predicted value of each of the user-level video data indicators, the method further includes:

[0014] Acquire historical video service DPI data of the multiple users from the transport layer and the network layer of the network model, and acquire N historical user-level video data indicators of the multiple users from the application layer of the network model;

[0015] For each of the N historical user-level video data indicators, based on the historical user-level video data indicators of the multiple users, feature screening is performed on all historical video service DPI indicators in the historical video service DPI data of the multiple users by using the Pearson correlation coefficient to obtain historical target DPI data with high correlation with the historical user-level video data indicators;

[0016] Taking each of the historical user-level video data indicators of the multiple users as a target variable, taking the historical target DPI data of the multiple users with high correlation with each of the historical user-level video data indicators as an independent variable, to obtain N target variables and corresponding independent variables;

[0017] For each of the target variables and the corresponding independent variables, a corresponding indicator prediction model is constructed and trained according to the target variable and the corresponding independent variable to obtain N trained indicator prediction models.

[0018] Further, according to the historical user-level video data indicators of the multiple users, feature screening is performed on all historical video service DPI indicators in the historical video service DPI data of the multiple users by using the Pearson correlation coefficient to obtain historical target DPI data with high correlation with the historical user-level video data indicators, specifically including:

[0019] For each of the N historical user-level video data indicators, the Pearson correlation coefficients of the historical user-level video data indicators of the multiple users and all the historical video service DPI indicators of the multiple users are calculated respectively, and the historical video service DPI indicators whose Pearson correlation coefficients are greater than a preset threshold are screened out as historical target DPI data with a high correlation with the historical user-level video data indicators.

[0020] Further, before performing feature screening on all historical video service DPI indicators in the historical video service DPI data of the multiple users by the Pearson correlation coefficient according to the historical user-level video data indicators of the multiple users to obtain historical target DPI data with a high correlation with the historical user-level video data indicators, the method further includes:

[0021] Data cleaning and outlier processing are performed on the historical video service DPI data of the multiple users and N historical user-level video data indicators.

[0022] Furthermore, the N user-level video data indicators include at least two of the following: number of video playbacks, number of successful video playbacks, video playback success rate, number of playback freezes, playback freeze rate, total first frame loading time, average first frame loading time, video downlink traffic, video downlink time, video download rate, total video time, video freeze time, and video freeze time ratio.

[0023] Furthermore, the indicator prediction model is an extreme gradient boosting tree XGBOOST model.

[0024] In a second aspect, the present invention provides a device for obtaining user-level video data indicators, comprising:

[0025] An acquisition module is used to obtain the user's video service deep packet inspection DPI data;

[0026] A prediction module, connected to the acquisition module, is used to input the video service DPI data of the user into N trained indicator prediction models, and predict the N user-level video data indicators of the user through the N indicator prediction models, wherein the trained N indicator prediction models are trained based on historical video service DPI data of multiple users and N historical user-level video data indicators, and N is greater than or equal to 1.

[0027] In a third aspect, the present invention provides a user-level video data indicator acquisition device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the user-level video data indicator acquisition method described in the first aspect above.

[0028] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for obtaining user-level video data indicators described in the first aspect is implemented.

[0029] The user-level video data indicator acquisition method, device and medium provided by the present invention first acquire the user's video service deep packet inspection DPI data; then input the user's video service DPI data into N trained indicator prediction models, and predict the user's N user-level video data indicators through the N indicator prediction models, wherein the trained N indicator prediction models are trained based on historical video service DPI data of multiple users and N historical user-level video data indicators, and N is greater than or equal to 1. The present invention predicts the user's N user-level video data indicators through the user's video service DPI data, using N indicator prediction models trained based on historical video service DPI data of multiple users and N historical user-level video data indicators, thereby greatly reducing the difficulty of acquiring user-level video data indicators, and solving the problem that the existing user-level video data indicator acquisition method is limited by base station manufacturers, resulting in difficulty in acquiring indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flow chart of a method for obtaining user-level video data indicators according to Embodiment 1 of the present invention;

[0031] Figure 2 A flowchart of another method for obtaining user-level video data indicators according to an embodiment of the present invention;

[0032] Figure 3 This is a structural diagram of a device for obtaining user-level video data indicators according to Embodiment 2 of the present invention;

[0033] Figure 4 This is a structural diagram of a device for obtaining user-level video data indicators according to Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0034] In order to enable those skilled in the art to better understand the technical solution of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0035] It should be understood that the specific embodiments and drawings described herein are only used to explain the present invention rather than to limit the present invention.

[0036] It can be understood that, in the absence of conflict, the various embodiments of the present invention and the various features in the embodiments can be combined with each other.

[0037] It can be understood that, for the convenience of description, the drawings of the present invention only show the parts related to the present invention, while the parts irrelevant to the present invention are not shown in the drawings.

[0038] It can be understood that each unit and module involved in the embodiments of the present invention may correspond to only one physical structure, or may be composed of multiple physical structures, or multiple units and modules may be integrated into one physical structure.

[0039] It can be understood that, without conflict, the functions and steps marked in the flowcharts and block diagrams of the present invention may occur in an order different from that marked in the drawings.

[0040] It is understood that the flowcharts and block diagrams of the present invention illustrate the possible architectures, functions, and operations of the systems, devices, equipment, and methods according to the various embodiments of the present invention. Each box in the flowchart or block diagram may represent a unit, module, program segment, or code, which contains executable instructions for implementing the specified functions. Moreover, each box or combination of boxes in the block diagram and flowchart may be implemented by a hardware-based system that implements the specified functions, or may be implemented by a combination of hardware and computer instructions.

[0041] It can be understood that the units and modules involved in the embodiments of the present invention can be implemented by software or hardware. For example, the units and modules can be located in a processor.

[0042] Application Overview

[0043] The existing method for obtaining user-level video data indicators directly collects user-level video data indicators from the core network through professional network management to build an indicator evaluation system.

[0044] However, the disadvantage of this method is that the definition and collection scope of user-level video data indicators are limited by base station manufacturers, which makes it difficult to obtain indicators. It is even possible that only a small part of the indicators can be collected and associated, which cannot support subsequent user perception and quality optimization analysis.

[0045] In response to the above technical problems, the present application provides a method, device and medium for obtaining user-level video data indicators. Through the user's video service DPI data, N indicator prediction models obtained by training based on historical video service DPI data of multiple users and N historical user-level video data indicators are used to predict the user's N user-level video data indicators, thereby greatly reducing the difficulty of obtaining user-level video data indicators, so as to at least solve the problem that the existing user-level video data indicator acquisition method is limited by base station manufacturers, resulting in difficulty in obtaining indicators.

[0046] After introducing the basic principles of the present application, various non-limiting embodiments of the present application will be described in detail with reference to the accompanying drawings.

[0047] Embodiment 1:

[0048] This embodiment provides a method for obtaining user-level video data indicators, such as Figure 1 As shown, the method includes:

[0049] Step S101: Acquire the user's video service DPI (Deep Packet Inspection) data.

[0050] It should be noted that the video service deep packet inspection DPI data may include: average DNS (Domain Name System) query time, TCP (Transmission Control Protocol) connection request times, TCP successful establishment times, GET request times, GET response times, play request times, play successful times, TCP establishment time, GET response time, initial buffering time, play time, pause times, streaming media service report times, downlink TCP retransmission packet number, downlink TCP total packet number, streaming media packet download volume, streaming media packet download time, pause time, video streaming media playback interruption times, video streaming media playback times, RST (Reset) interruption times, FIN (Finish) interruption times, streaming media playback waiting times, streaming media pause frequency, streaming media pause proportion The number of times the ratio is too high, the number of times the video streaming download rate is low, the number of times the VAP streaming playback wait time is too long, the number of times the VAP streaming pause frequency is too high, the number of times the VAP streaming pause ratio is too high, the number of times the VAP video streaming download rate is low, the number of video streaming GET failures, the number of video streaming initial buffering failures, the number of video streaming playback interruptions, the number of streaming playback interruptions including user failures, the number of GET responses including user failures, the number of video streaming xKB startup delays too long, the number of video streaming encryption services, the total xKB startup delay, the number of xKB starts, the total downlink RTT (Round-TripTime), the number of downlink RTT calculations, etc.

[0051] Step S102: Input the video service DPI data of the user into N trained indicator prediction models, and predict N user-level video data indicators of the user through the N indicator prediction models, wherein the trained N indicator prediction models are trained based on historical video service DPI data of multiple users and N historical user-level video data indicators, and N is greater than or equal to 1.

[0052] In this embodiment, the indicator prediction model is preferably an extreme gradient boosting tree XGBOOST model, and the N user-level video data indicators include at least two of the following: number of video plays, number of successful video plays, video play success rate, number of playback freezes, playback freeze rate, total first frame loading time, average first frame loading time, video downstream traffic, video downstream time, video download rate, total video time, video freeze time, and video freeze time ratio.

[0053] In an optional embodiment, obtaining the deep packet inspection (DPI) data of the user's video service specifically includes:

[0054] Obtaining the video service DPI data from the transport layer and the network layer of the network model;

[0055] The step of inputting the video service DPI data of the user into the trained N indicator prediction models, and predicting the N user-level video data indicators of the user by using the N indicator prediction models, specifically includes:

[0056] Extracting target DPI data having a high correlation with each of the user-level video data indicators from the acquired video service DPI data;

[0057] The target DPI data with a high correlation with each of the user-level video data indicators is respectively input into the corresponding trained indicator prediction model to obtain the predicted value of each of the user-level video data indicators.

[0058] Specifically, the user's video service DPI data is collected from the transport layer and network layer of the network model, and for each user-level video data indicator, the target DPI data with a high correlation with the user-level video data indicator is extracted from the collected user's video service DPI data, and the target DPI data with a high correlation with each user-level video data indicator is respectively input into the corresponding trained indicator prediction models in N indicator prediction models to obtain the predicted values ​​of the user's N user-level video data indicators.

[0059] In an optional embodiment, before inputting the target DPI data having a high correlation with each of the user-level video data indicators into the corresponding trained indicator prediction model to obtain the predicted value of each of the user-level video data indicators, the method further includes:

[0060] Acquire historical video service DPI data of the multiple users from the transport layer and the network layer of the network model, and acquire N historical user-level video data indicators of the multiple users from the application layer of the network model;

[0061] For each of the N historical user-level video data indicators, based on the historical user-level video data indicators of the multiple users, feature screening is performed on all historical video service DPI indicators in the historical video service DPI data of the multiple users by using the Pearson correlation coefficient to obtain historical target DPI data with high correlation with the historical user-level video data indicators;

[0062] Taking each of the historical user-level video data indicators of the multiple users as a target variable, taking the historical target DPI data of the multiple users with high correlation with each of the historical user-level video data indicators as an independent variable, to obtain N target variables and corresponding independent variables;

[0063] For each of the target variables and the corresponding independent variables, a corresponding indicator prediction model is constructed and trained according to the target variable and the corresponding independent variable to obtain N trained indicator prediction models.

[0064] Specifically, historical video service DPI data of multiple users are collected from the transport layer and the network layer of the network model, and N historical user-level video data indicators of multiple users are collected from the application layer of the network model.

[0065] Specifically, for each of the N historical user-level video data indicators, the Pearson correlation coefficients of the historical user-level video data indicators of the multiple users and all the historical video service DPI indicators of the multiple users are calculated respectively, and the historical video service DPI indicators whose Pearson correlation coefficients are greater than a preset threshold are screened out as historical target DPI data with a high correlation with the historical user-level video data indicators.

[0066] Specifically, an indicator prediction model is established for each user-level video data indicator, and each historical user-level video data indicator of multiple users is used as the target variable, and the historical target DPI data of multiple users with high correlation with each historical user-level video data indicator is used as an independent variable. Each target variable and the corresponding independent variable are input into the corresponding indicator prediction model for training to obtain N trained indicator prediction models.

[0067] It should be noted that the trained indicator prediction model can be evaluated by the mean square error (MSE) and the mean absolute error (MAE).

[0068] In an optional embodiment, before performing feature screening on all historical video service DPI indicators in the historical video service DPI data of the multiple users by using the Pearson correlation coefficient according to the historical user-level video data indicators of the multiple users to obtain historical target DPI data with a high correlation with the historical user-level video data indicators, the method further includes:

[0069] Data cleaning and outlier processing are performed on the historical video service DPI data of the multiple users and N historical user-level video data indicators.

[0070] Specifically, the historical video service DPI data of multiple users and N historical user-level video data indicators are cleaned: the indicators with missing values ​​exceeding 90% and the indicators with zero values ​​exceeding 80% are removed. If there is a missing value in any field of the historical video service DPI data and N historical user-level video data indicators of a certain user, the data will not be included in the training; the historical video service DPI data and N historical user-level video data indicators of multiple users are detected for outliers, and the historical video service DPI data and N historical user-level video data indicators of the users containing the outliers are deleted.

[0071] In a specific embodiment, a feature screening algorithm based on XGBOOST regression tree and Pearson correlation coefficient is used, with XGBOOST regression tree as the main prediction model, and DPI data is screened by Pearson correlation coefficient, such as Figure 2 As shown, the overall steps of the method for obtaining user-level video data indicators are:

[0072] S1. Collect user-level video data of multiple users in the past six months from the application layer of the network model, and collect DPI data of multiple users in the past six months from the transport layer and network layer of the network model

[0073] Specifically, the network model is the TCP / IP (Transmission Control Protocol / Internet Protocol) network model, and video-level data is collected from the application layer of the network model, including: video play times (PLAY_CNT), video play success times (PLAY_SUCC_CNT), video play success rate (PLAY_SUCC_RATE), play freeze times (CATON_CNT), play freeze rate (CATON_RATE), total first frame loading time (LOAD_SUM_DUR), average first frame loading time (LOAD_AVG_DUR), video downstream traffic (DOWN_FLOW), video downstream time (DOWN_DUR), video download rate (DOWN_RATE), total video time (PLAY_DUR), video freeze time (CATON_DUR), and video freeze time ratio (CATON_DUR_RAT). DPI data is collected from the transport layer and network layer of the network model, including: DNS Query Delay, TCP Connection Requests, TCP Connection Success Times, GET Requests, GET Responses, Streaming GET Success Times, Streaming Initial Buffer Requests, Video Streaming Requests, Streaming Initial Buffer Success Times, Video Streaming Successes, TCP Connect Delay, GET Response Delay, Initial Buffer Delay, Streaming Play Duration, Stall Count, Streaming Service Report Frequency, TCP Download Retransmission Packets, TCP Download Packets, Streaming Download Packets, Video Streaming Requests, Streaming Initial Buffer Success Times, TCP Connection Delay, GET Response Delay, Initial Buffer Delay, Streaming Play Duration, Stall Count, Streaming Service Report Frequency, TCP Download Retransmission Packets, TCP Download Packets, Streaming Download PacketsPackets), Streaming Download Delay, Stall Duration, Video Streaming Disconnections, Video Streaming Plays, RST Disconnections, FIN Disconnections, Streaming Response Long Delay Times, Streaming High Stall Frequency Times, Streaming High Stalled Time Rate Times, Streaming Low Download Throughput Times, VAP Streaming Response Long Delay Times, VAP Streaming High Stall Frequency Times, VAP Streaming High Stalled Time Rate Times, VAP Video Streaming Download Low Times, Video Streaming GET Failures (User Reasons) (Video Streaming Failures(User), Video Streaming Initial Buffer Failures(User), Video Streaming Disconnections(User), Video Streaming Disconnection Num Include User, Video Get Response Num Include User, Video Streaming xKB Start Delays(VideoStreaming xKB Start Delays), Video Streaming Encrypted Services(Video Streaming Disconnections(User), Video Streaming Disconnection Num Include User, Video Streaming Response Num Include User, Video Streaming xKB Start Delays(Video Streaming xKB Start Delays), Video Streaming Encrypted Services(Video Streaming Disconnections(User), Video Streaming Disconnection Num Include User, ...The fields include ServiceTimes, xKB Start Delay, xKB Start Times, Downlink Total RTT, and Downlink RTT Count, totaling 41 fields.

[0074] S2. Clean user-level video data and DPI data (null value processing)

[0075] Specifically, remove data with missing values ​​exceeding 90%. If the missing values ​​of a field exceed 90%, delete all data of the field. Remove data with zero values ​​exceeding 80%. If the zero values ​​of a field exceed 80%, delete all data of the field. If there is a missing value in any field of the user-level video data and the corresponding DPI data of a user, the data will not be included in the training.

[0076] S3. Perform outlier detection and delete user-level video data and DPI data containing abnormal data (error value processing)

[0077] Specifically, outlier detection is performed, and the data of each indicator is input into the isolation forest model for model training. After the model training is completed, the data of each indicator is input into the model again for outlier detection. If the number of abnormal fields in the video data and DPI data of a user exceeds a, the data of the user is considered to be abnormal and the user data is deleted. After experiments and expert verification, the anomaly detection result is most accurate when a is 10.

[0078] S4. Calculate the Pearson correlation coefficient between each user-level video data indicator and the full amount of DPI data, and remove DPI indicators with correlation coefficients less than 0.1 (low correlation data cleaning)

[0079] Specifically, the Pearson correlation coefficient of each user-level video data indicator and the full DPI data is calculated, where the Pearson correlation coefficient is widely used to measure the correlation between two variables, and the value of the correlation coefficient is between -1 and 1. -1 indicates that the two variables are completely negatively correlated, and 1 indicates that the two variables are completely positively correlated. The calculation method is shown in formula (1):

[0080]

[0081] Among them, ρ X,Y represents the Pearson correlation coefficient between variable X and variable Y, σ X represents the variance of variable X, σ Y represents the variance of variable Y, cov(X,Y) represents the covariance of variables X and Y, μX represents the mean of variable X, μ Y represents the mean of the variable Y, and E[ ] represents the operation of obtaining the expectation. Because the DPI index with a high correlation coefficient with the user-level video data index is more important, it can make the model converge faster and improve the prediction accuracy of the regression prediction model (that is, the index prediction model). Therefore, after the calculation, the data of the DPI index with a correlation coefficient less than b is removed. After experiments and manual verification by experts, when b is 0.1, the DPI data index obtained by screening has the highest improvement on the model accuracy.

[0082] S5. Preprocessing of DPI data for feature screening

[0083] Specifically, the z-score method is used for standardization to reduce the impact of the dimension of the indicator on the regression prediction of user-level video data, and the standardized data will be used for regression prediction. Secondly, the video data after feature screening and the high-correlation DPI data are divided in a ratio of 8:2, with 80% of the data used for training and 20% for testing.

[0084] S6. Build an XGBOOST regression tree model for each user-level video data indicator

[0085] Specifically, use training data. Import the XGBOOST library, initialize the regression model, set parameters such as the tree depth, learning rate, and number of iterations, and adjust and optimize based on experience or multiple experiments. Use the training set data, take the user-level video data index as the target (that is, the target variable), and the corresponding DPI data index as the feature (that is, the independent variable), and input it into the XGBOOST model for training. The model will build multiple regression trees through continuous iterations, and optimize the parameters of the model based on the gradient descent direction of the loss function, so that the error between the predicted value and the actual target value gradually decreases. Use the test set data to evaluate the trained model with mean square error (MSE), mean absolute error (MAE), etc. to measure the prediction accuracy and performance of the model. According to the evaluation results, if the model performance does not meet expectations, the model parameters can be adjusted, such as increasing the number of trees, adjusting the learning rate, etc., and then retraining and evaluating until satisfactory model performance is obtained.

[0086] Specifically, for each user-level video data indicator, an XGBOOST regression prediction model is established, and the learning rate (learning_rate) is set to 0.05, the number of trees (n_estimators) is set to 100, the maximum depth (max_depth) is set to 7, the minimum loss reduction value (gamma) of node splitting is set to 0.1, the random sampling ratio of the number of columns of each tree (colsample_bytree) is set to 0.9, the random sampling ratio of the number of samples of each tree (subsample) is set to 0.9, and the L1 regularization coefficient (reg_alpha) is set to 0.005. The model is trained using the training data. After the training, the model is tested using the test data. Adjust the above parameters according to the test results to optimize the model. Finally, the model is persisted locally.

[0087] S7. Use the trained model to predict user-level video data indicators based on a large amount of DPI data

[0088] Specifically, multiple XGBOOST regression prediction models are used to predict multiple user-level video data indicators based on a large amount of DPI data, and the predicted values ​​of multiple user-level video data indicators and the names of the DPI indicators with the TOP10 feature importance corresponding to each video data indicator are output.

[0089] It is worth mentioning that the present invention uses the Pearson correlation coefficient to perform feature screening on a large number of DPI data indicators, establishes the relationship between DPI data and user-level video data indicators, takes each user-level video data indicator as the target variable, and uses the data of the corresponding DPI indicator with higher correlation as the independent variable. The corresponding regression prediction model is trained for each user-level video data indicator, so as to accurately predict the user-level video data indicator through the regression prediction model. The present invention collects the user-level video data indicators of the user, combines DPI collection, associates the user-level service details, and introduces the AI ​​algorithm to realize the prediction of the user-level video data indicators of the user, so as to further analyze the user perception and quality. The present invention calculates the Pearson correlation coefficient of each user-level video data indicator and the full amount of DPI data, obtains the correlation relationship according to the correlation coefficient, and uses XGBOOST to model the relationship between DPI data and user-level video data indicators. A large amount of available DPI data can be used to predict the corresponding user-level video data indicators to support the subsequent analysis of user perception and quality optimization. The present invention uses the Pearson correlation coefficient for feature screening, and finds fields with high correlation with user-level video data indicators in a large number of DPI fields for model training. The speed of model convergence and the accuracy of model prediction are improved.

[0090] The user-level video data indicator acquisition method provided by the embodiment of the present invention first acquires the user's video service deep packet inspection DPI data; then the user's video service DPI data is input into the trained N indicator prediction models, and the N user-level video data indicators of the user are predicted by the N indicator prediction models, wherein the trained N indicator prediction models are trained based on the historical video service DPI data of multiple users and N historical user-level video data indicators, and N is greater than or equal to 1. The present invention predicts the user's N user-level video data indicators through the user's video service DPI data, using the N indicator prediction models trained based on the historical video service DPI data of multiple users and N historical user-level video data indicators, thereby greatly reducing the difficulty of acquiring user-level video data indicators, and solving the problem that the existing user-level video data indicator acquisition method is limited by base station manufacturers, resulting in difficulty in acquiring indicators.

[0091] Embodiment 2:

[0092] like Figure 3 As shown, this embodiment provides a user-level video data indicator acquisition device, which is used to execute the above-mentioned user-level video data indicator acquisition method, including:

[0093] An acquisition module 11 is used to acquire the user's video service deep packet inspection DPI data;

[0094] The prediction module 12 is connected to the acquisition module 11, and is used to input the video service DPI data of the user into N trained indicator prediction models, and predict the N user-level video data indicators of the user through the N indicator prediction models, wherein the trained N indicator prediction models are trained based on historical video service DPI data of multiple users and N historical user-level video data indicators, and N is greater than or equal to 1.

[0095] Furthermore, the acquisition module 11 specifically includes:

[0096] A first acquisition unit, configured to acquire the video service DPI data from a transport layer and a network layer of a network model;

[0097] The prediction module 12 specifically includes:

[0098] An extraction unit, configured to extract target DPI data having a high correlation with each of the user-level video data indicators from the acquired video service DPI data;

[0099] The prediction unit is used to input the target DPI data with high correlation with each of the user-level video data indicators into the corresponding trained indicator prediction model to obtain the predicted value of each of the user-level video data indicators.

[0100] Furthermore, the prediction module 12 also includes:

[0101] A second acquisition unit is used to acquire the historical video service DPI data of the multiple users from the transport layer and the network layer of the network model, and acquire N historical user-level video data indicators of the multiple users from the application layer of the network model;

[0102] A screening unit is used for performing feature screening on all historical video service DPI indicators in the historical video service DPI data of the multiple users according to the historical user-level video data indicators of the multiple users by using the Pearson correlation coefficient for each of the historical user-level video data indicators in the N historical user-level video data indicators, so as to obtain historical target DPI data with high correlation with the historical user-level video data indicators;

[0103] As a unit, used to respectively use each of the historical user-level video data indicators of the multiple users as a target variable, and use the historical target DPI data of the multiple users with high correlation with each of the historical user-level video data indicators as an independent variable, to obtain N target variables and corresponding independent variables;

[0104] A training unit is constructed to construct and train a corresponding indicator prediction model for each target variable and the corresponding independent variable according to the target variable and the corresponding independent variable, so as to obtain N trained indicator prediction models.

[0105] Furthermore, the screening unit is specifically used for:

[0106] For each of the N historical user-level video data indicators, the Pearson correlation coefficients of the historical user-level video data indicators of the multiple users and all the historical video service DPI indicators of the multiple users are calculated respectively, and the historical video service DPI indicators whose Pearson correlation coefficients are greater than a preset threshold are screened out as historical target DPI data with a high correlation with the historical user-level video data indicators.

[0107] Furthermore, the prediction module 12 also includes:

[0108] The cleaning processing unit is used to perform data cleaning and outlier processing on the historical video service DPI data of the multiple users and N historical user-level video data indicators.

[0109] Furthermore, the N user-level video data indicators include at least two of the following: number of video playbacks, number of successful video playbacks, video playback success rate, number of playback freezes, playback freeze rate, total first frame loading time, average first frame loading time, video downlink traffic, video downlink time, video download rate, total video time, video freeze time, and video freeze time ratio.

[0110] Furthermore, the indicator prediction model is an extreme gradient boosting tree XGBOOST model.

[0111] Embodiment 3:

[0112] refer to Figure 4 This embodiment provides a user-level video data indicator acquisition device, including a memory 21 and a processor 22. The memory 21 stores a computer program, and the processor 22 is configured to run the computer program to execute the user-level video data indicator acquisition method in Example 1.

[0113] The memory 21 is connected to the processor 22. The memory 21 may be a flash memory, a read-only memory or other memory. The processor 22 may be a central processing unit or a single-chip microcomputer.

[0114] Embodiment 4:

[0115] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for obtaining a user-level video data indicator in the above-mentioned embodiment 1 is implemented.

[0116] The computer-readable storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). Computer-readable storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable read only memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0117] In summary, the user-level video data indicator acquisition method, device and medium provided by the embodiments of the present invention first acquire the user's video service deep packet inspection DPI data; then input the user's video service DPI data into the trained N indicator prediction models, and predict the user's N user-level video data indicators through the N indicator prediction models, wherein the trained N indicator prediction models are trained based on the historical video service DPI data of multiple users and N historical user-level video data indicators, and N is greater than or equal to 1. The present invention predicts the user's N user-level video data indicators through the user's video service DPI data, using N indicator prediction models trained based on the historical video service DPI data of multiple users and N historical user-level video data indicators, thereby greatly reducing the difficulty of acquiring user-level video data indicators, and solving the problem that the existing user-level video data indicator acquisition method is limited by base station manufacturers, resulting in difficulty in acquiring indicators.

[0118] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for obtaining user-level video data indicators, characterized in that: The method comprises: Obtain the user's video service deep packet inspection DPI data; The video service DPI data of the user is input into N trained indicator prediction models, and N user-level video data indicators of the user are predicted by the N indicator prediction models, wherein the trained N indicator prediction models are trained based on historical video service DPI data of multiple users and N historical user-level video data indicators, and N is greater than or equal to 1.

2. The method according to claim 1, characterized in that The obtaining of the user's video service deep packet inspection DPI data specifically includes: Obtaining the video service DPI data from the transport layer and the network layer of the network model; The step of inputting the video service DPI data of the user into the trained N indicator prediction models, and predicting the N user-level video data indicators of the user by using the N indicator prediction models, specifically includes: Extracting target DPI data having a high correlation with each of the user-level video data indicators from the acquired video service DPI data; The target DPI data with a high correlation with each of the user-level video data indicators is respectively input into the corresponding trained indicator prediction model to obtain the predicted value of each of the user-level video data indicators.

3. The method according to claim 2, characterized in that Before inputting the target DPI data with high correlation with each of the user-level video data indicators into the corresponding trained indicator prediction model to obtain the predicted value of each of the user-level video data indicators, the method further includes: Acquire historical video service DPI data of the multiple users from the transport layer and the network layer of the network model, and acquire N historical user-level video data indicators of the multiple users from the application layer of the network model; For each of the N historical user-level video data indicators, based on the historical user-level video data indicators of the multiple users, feature screening is performed on all historical video service DPI indicators in the historical video service DPI data of the multiple users by using the Pearson correlation coefficient to obtain historical target DPI data with high correlation with the historical user-level video data indicators; Taking each of the historical user-level video data indicators of the multiple users as a target variable, taking the historical target DPI data of the multiple users with high correlation with each of the historical user-level video data indicators as an independent variable, to obtain N target variables and corresponding independent variables; For each of the target variables and the corresponding independent variables, a corresponding indicator prediction model is constructed and trained according to the target variable and the corresponding independent variable to obtain N trained indicator prediction models.

4. The method according to claim 3, characterized in that The method of performing feature screening on all historical video service DPI indicators in the historical video service DPI data of the multiple users by using the Pearson correlation coefficient according to the historical user-level video data indicators of the multiple users to obtain historical target DPI data with high correlation with the historical user-level video data indicators specifically includes: For each of the N historical user-level video data indicators, the Pearson correlation coefficients of the historical user-level video data indicators of the multiple users and all the historical video service DPI indicators of the multiple users are calculated respectively, and the historical video service DPI indicators whose Pearson correlation coefficients are greater than a preset threshold are screened out as historical target DPI data with a high correlation with the historical user-level video data indicators.

5. The method according to claim 3, characterized in that: Before performing feature screening on all historical video service DPI indicators in the historical video service DPI data of the multiple users by using the Pearson correlation coefficient according to the historical user-level video data indicators of the multiple users to obtain historical target DPI data with high correlation with the historical user-level video data indicators, the method further includes: Data cleaning and outlier processing are performed on the historical video service DPI data of the multiple users and N historical user-level video data indicators.

6. The method according to claim 1, characterized in that The N user-level video data indicators include at least two of the following: number of video play times, number of successful video play times, video play success rate, number of playback freezes, playback freeze rate, total first frame loading time, average first frame loading time, video downlink traffic, video downlink time, video download rate, total video time, video freeze time, and video freeze time ratio.

7. The method according to claim 1, characterized in that The indicator prediction model is an extreme gradient boosting tree XGBOOST model.

8. A device for obtaining user-level video data indicators, characterized in that: include: An acquisition module is used to obtain the user's video service deep packet inspection DPI data; A prediction module, connected to the acquisition module, is used to input the video service DPI data of the user into N trained indicator prediction models, and predict the N user-level video data indicators of the user through the N indicator prediction models, wherein the trained N indicator prediction models are trained based on historical video service DPI data of multiple users and N historical user-level video data indicators, and N is greater than or equal to 1.

9. A device for obtaining user-level video data indicators, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to implement the method for obtaining user-level video data indicators according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for obtaining a user-level video data indicator according to any one of claims 1 to 7 is implemented.

Images