A live video data intelligent processing system and method

Through the combination of crawling technology and machine learning algorithms, key indicators in live broadcast data are collected and evaluated, and the live broadcast effect evaluation indicator U is generated, which solves the problem of the inability to comprehensively evaluate anchor interaction and participation in the existing technology, and achieves more accurate live broadcast Effectiveness evaluation and user experience improvement.

CN118474426BActive Publication Date: 2024-11-22GUANGZHOU HELMSMAN NETWORK TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410675772.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-29
Publication Date
2024-11-22
Estimated Expiration
2044-05-29

AI Technical Summary

Technical Problem

The existing live video data processing methods fail to comprehensively evaluate the anchor's interaction and participation, ignore the number of barrage sent and the frequency of interaction between the anchor and the audience, and cannot accurately reflect the anchor's live broadcast effect and popularity, affecting the user experience.

Method used

Use crawler technology to collect live broadcast data, automatically classify and preprocess it through machine learning algorithms, and combine the number of viewers, the number of barrage sends, the frequency of interaction between anchor and audience, and the number of real-time violations to generate live broadcast effect evaluation indicator U, which is used for sorting and pushing high-quality Anchor.

Benefits of technology

Provide more comprehensive and multi-dimensional evaluation indicators, improve the real-time and flexibility of evaluation, accurately reflect the live broadcast effect and popularity of anchors, improve user viewing and increase the traffic of anchors and platforms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118474426B_ABST
    Figure CN118474426B_ABST
Patent Text Reader

Abstract

The invention discloses a live video data intelligent processing system and method, and relates to the technical field of live data processing. The system comprises an information collection module, a pre-classification module, an evaluation and analysis module, and an intelligent push module. The technical key points are as follows: by comprehensively considering multiple evaluation parameters such as the number of viewers, the amount of barrage sent, the frequency of interaction between the anchor and the audience, etc., more comprehensive and multi-dimensional evaluation indicators can be provided. When obtaining the evaluation indicators, the number of real-time violations of the anchor is also considered at the same time, and the occurrence of violations is measured by the fluctuation of the number of viewers. The violations of the anchor can be observed more carefully and the violations can be monitored in time. The evaluation indicators can also more accurately reflect the live broadcast effect and popularity of the anchor, so that when users search for anchors of this type, they can push anchors with higher quality, improve the user's viewing experience, and increase continuous traffic for the anchor and the platform.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of live data processing, and in particular to a live video data intelligent processing system and method. Background Art

[0002] Live broadcast data refers to various live broadcast-related information and indicator data collected during the live broadcast process, usually including anchor information, live broadcast content, number of viewers, amount of barrage sent, and other data. When processing live broadcast data, the following steps are usually required: Data collection: Collect data generated during the live broadcast process in real time or regularly through the API interface of the live broadcast platform or other means; Data cleaning: Clean and pre-process the collected data to remove duplicate data, process missing values, correct erroneous data, etc., to ensure the reliability and accuracy of the data; Data storage: Store the cleaned data in a database or other storage system for subsequent analysis and application; Data analysis: Analyze the data through statistical methods to extract valuable information; Result presentation: Visually display the results of the analysis, such as displaying the changing trends and statistical analysis results of indicators such as the number of viewers, amount of barrage sent, and frequency of interaction of the anchor in the form of charts and reports.

[0003] The technical solution pointed out in the document of the existing application publication number CN105306966A and titled A method, device and system for processing live video data includes: an HTTP server based on a Linux platform in a server responds to a first live broadcast request sent by a client and carrying a video bit rate and a live broadcast program source, and sends an m3u8 file corresponding to the video bit rate and the live broadcast program source to the client; and, in response to a second live broadcast request sent by the client and carrying a video index name obtained from the m3u8 file, obtains a memory address of a TS segment corresponding to the video index name from a preset hash table; and obtains a TS segment from a memory corresponding to the memory address of the TS segment; and sends the TS segment to the client that sends the second live broadcast request; therefore, the present application saves disk space, improves the efficiency of responding to user requests, and improves user experience, but does not evaluate the situation of the corresponding anchor by comprehensively processing the live broadcast data.

[0004] In combination with the above documents and existing technologies, traditionally designed processing methods for live video data usually only consider basic indicators such as the number of viewers, without comprehensively considering other important indicators such as the number of barrages sent, the frequency of interaction between the anchor and the audience, etc. This will ignore the interactivity and participation of the live broadcast, and cannot comprehensively evaluate the live broadcast effect and the popularity of the anchor. Even if some indicator data are considered, the currently designed processing system does not comprehensively consider the real-time number of violations of the anchor in the scheme for evaluating the live broadcast effect of the anchor. It is only used to monitor the anchor's violations. When the user searches for the corresponding type of game, the platform recommends several anchors. The live broadcast effects of these anchors are different. If the user enters to watch the anchor with poor live broadcast effect, it will greatly reduce the user's interest, thereby affecting the user's viewing experience. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the shortcomings of the prior art, the present invention provides a system and method for intelligent processing of live video data. By comprehensively considering multiple evaluation parameters such as the number of viewers, the amount of barrage sent, the frequency of interaction between the anchor and the audience, etc., it can provide more comprehensive and multi-dimensional evaluation indicators. When obtaining the evaluation indicators, the number of real-time violations of the anchor is also considered, and the occurrence of violations is measured by the fluctuation of the number of viewers. This method makes the evaluation more real-time and flexible, and can observe the violations of the anchor more carefully and monitor the violations in time. It can also more accurately reflect the live broadcast effect and popularity of the anchor through the evaluation indicator, that is, the live broadcast effect evaluation index U, so that users can push higher-quality anchors when searching for this type of anchors, which improves the user's viewing experience while increasing continuous traffic for the anchor and the platform, thereby solving the problems raised in the background technology.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A live video data intelligent processing system, comprising:

[0010] The information collection module uses crawler technology to collect live video data from live data sources;

[0011] The pre-classification module automatically classifies the collected live video data by using machine learning algorithms;

[0012] The evaluation and analysis module matches the corresponding live broadcast type according to the search content, obtains the required information in the live broadcast video data of the corresponding anchor under the live broadcast type, and extracts evaluation parameters from the required information of each corresponding type of anchor for a second time. The evaluation parameters include the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience. Based on the pre-processed evaluation parameters and the real-time number of violations and live broadcast duration of the corresponding type of anchor, a live broadcast data evaluation model is built to generate the live broadcast effect evaluation value Ir of the corresponding anchor;

[0013] The intelligent push module sorts the anchors corresponding to the same type of live broadcast data according to the size of the live broadcast effect evaluation index U, and generates a sorting table. The live broadcast effect evaluation index U in the sorting table is arranged from large to small. When searching for anchors of this type, the corresponding anchors are pushed in sequence according to the sorting table.

[0014] Furthermore, the live broadcast data source includes various live broadcast platforms and websites, and the collected live broadcast video data includes the required information, which includes anchor information, live broadcast room information, number of viewers, barrage content and interactive information.

[0015] Furthermore, the anchor information includes anchor ID, anchor nickname, and anchor level; the live broadcast room information includes live broadcast room ID, live broadcast room title, and live broadcast duration; the number of viewers includes the current number of viewers and the peak number of viewers; the barrage content includes the nickname of the barrage sender, the barrage sentence, the barrage sending time, and the amount of barrage sent per unit time; the interactive information includes the frequency of interaction between the anchor and the audience.

[0016] Furthermore, the machine learning algorithm used in the pre-classification module includes any one of a naive Bayes classifier, a support vector machine SVM, and a convolutional neural network CNN, and the pre-classification module is used to distinguish the game type to which the video belongs in the same type of game anchor video data.

[0017] Furthermore, the process of preprocessing the evaluation parameters, the number of real-time violations, and the live broadcast duration is: data cleaning and dimensionless processing of each data.

[0018] Furthermore, the process of obtaining the real-time number of violations of the corresponding type of anchor is as follows: during the normal live broadcast process, the number of times the fluctuation value of the number of viewers of the live video at adjacent moments exceeds the preset standard threshold F1, and the number of times the corresponding anchor has illegal content in the live broadcast room and received warnings from the administrator F2. By adding F1 and F2, the total obtained is the real-time number of violations of the corresponding type of anchor, among which the calculation method of the fluctuation value at adjacent moments is: the fluctuation value at adjacent moments = the number of online viewers at the previous moment × the number of online viewers at the next moment.

[0019] Further, the process of generating the live broadcast effect evaluation value Ir is as follows: S301, based on the current number of viewers, the amount of barrage sent per unit time, the frequency of interaction between the anchor and the audience, the number of real-time violations of the corresponding type of anchor, and the live broadcast duration, generate the live broadcast effect evaluation index U, and the formula based on it is as follows:

[0020]

[0021] In the formula (1), u Δ represents a positive index, gs, ds, and pc in formula (2) represent the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience, respectively, a1, a2, and a3 are the preset proportional coefficients of the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience, and a1>a2>a3>0, wc and zs in formula (3) represent the real-time violation times and live broadcast duration of the corresponding type of anchor, respectively, and G is a constant correction coefficient;

[0022] S302: Obtaining the live broadcast effect evaluation index U at different times within T time. t , t represents the number of the live broadcast effect evaluation index at different times within time T, t = 1, 2, 3, 4, ..., n, n is a positive integer;

[0023] S303: Based on the average value of the live broadcast effect evaluation index at different times within the time T and the live broadcast effect evaluation index U t , calculate the live broadcast effect evaluation value Ir, the calculation formula is as follows:

[0024]

[0025] In the formula, It represents the average value of the live broadcast effect evaluation index at different times within time T.

[0026] A method for intelligently processing live video data comprises the following steps:

[0027] S1. Use crawler technology to collect live video data from live data sources;

[0028] S2, automatically classifying the collected live video data by using machine learning algorithms;

[0029] S3. Match the corresponding live broadcast type according to the search content, and obtain the required information in the live broadcast video data of the corresponding anchor under the live broadcast type. Extract evaluation parameters from the required information of each corresponding type of anchor for the second time, and the evaluation parameters include the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience. According to the pre-processed evaluation parameters and the real-time number of violations and live broadcast duration of the corresponding type of anchor, build a live broadcast data evaluation model to generate the live broadcast effect evaluation value Ir of the corresponding anchor;

[0030] S4. Sort the anchors corresponding to the same type of live broadcast data according to the size of the live broadcast effect evaluation index U to generate a sorting table. The live broadcast effect evaluation index U in the sorting table is arranged from large to small. When searching for anchors of this type, push each corresponding anchor in turn according to the sorting table.

[0031] (III) Beneficial effects

[0032] The present invention provides a live video data intelligent processing system and method, which has the following beneficial effects:

[0033] 1. By comprehensively considering multiple evaluation parameters such as the number of viewers, the amount of barrage sent, the frequency of interaction between the anchor and the audience, it can provide more comprehensive and multi-dimensional evaluation indicators. When obtaining the evaluation indicators, the number of real-time violations of the anchor is also considered, and the occurrence of violations is measured by the fluctuation of the number of viewers. This method makes the evaluation more real-time and flexible, and can observe the anchor's violations in more detail and monitor violations in a timely manner. It can also more accurately reflect the anchor's live broadcast effect and popularity through the evaluation indicator, that is, the live broadcast effect evaluation index U;

[0034] 2. By processing the data generated during the live broadcast, different types of anchors in the game are evaluated, so that when users search for anchors of this type, higher-quality anchors can be pushed. While improving user viewing, it also increases continuous traffic for the anchor and the platform, achieving a mutually beneficial effect, which reflects the intelligent design of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a schematic diagram of the module structure of a live video data intelligent processing system in the present invention;

[0036] Figure 2 The figure is a schematic diagram of the overall process of a method for intelligently processing live video data in the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0038] Example 1: Please refer to Figure 1 This embodiment provides a live video data intelligent processing system. The application scenario of the system is for live broadcasting on computers or mobile phones. By processing the data generated during live broadcasting, it is convenient to evaluate various types of anchors, so that when users search for anchors of this type, they can push anchors with higher quality, which improves the user's viewing experience and increases continuous traffic for the anchor and the platform, achieving a mutually beneficial effect. At the same time, the system also processes the live video data generated by game anchors, and the processing system includes an information collection module, a pre-classification module, an evaluation and analysis module, and an intelligent push module;

[0039] The information collection module uses crawler technology to collect live video data from live data sources;

[0040] The live broadcast data source includes but is not limited to various live broadcast platforms and websites. The information collection module is used to collect live broadcast video data from various live broadcast platforms or websites. The data collection can be achieved through the API structure or by directly crawling the web page content.

[0041] Specific description: In the information collection module, the following steps can be used to collect a certain type of game live video data from the live broadcast platform: S101, determine the data source: select one or more live broadcast platforms, such as a TV, a live broadcast, etc. as the data source; S102, determine the collection object: determine to collect a certain type of game live video data, and filter the target according to the game classification or search keywords provided by the platform; S103, connect to the API interface or crawl the web page: obtain the live video data through the API interface provided by the live broadcast platform or directly crawl the web page content; S104, extract the required information: extract the anchor information, live broadcast room information, number of viewers, barrage content and interactive information from the collected data, which can be extracted by parsing the returned JSON or HTML data; S105, data storage: store the extracted live video data in a database or file to facilitate the use of subsequent data classification, evaluation and analysis modules;

[0042] For example, a certain type of game live video data collected from a TV may include the following information: anchor information: anchor ID, anchor nickname, anchor level; live broadcast room information: live broadcast room ID, live broadcast room title, live broadcast duration; number of viewers: current number of viewers, peak number of viewers; barrage content: barrage sender nickname, barrage sentence, barrage sending time, and barrage sending volume per unit time; interaction information: the frequency of interaction between the anchor and the audience; Through the above steps, we can realize the collection of a certain type of game live video data from the live broadcast platform and extract the required information for subsequent functional module processing.

[0043] The pre-classification module automatically classifies the collected live video data using a machine learning algorithm and aggregates video data of the same live type;

[0044] Among them, to classify the same type of game anchor video data, a text classification machine learning algorithm can be used, such as a naive Bayes classifier, a support vector machine SVM, and a convolutional neural network CNN. Therefore, the machine learning algorithm used in the pre-classification module includes any one of a naive Bayes classifier, a support vector machine SVM, and a convolutional neural network CNN.

[0045] The following is a specific explanation using the naive Bayes classifier as an example:

[0046] S201, data preprocessing: perform text preprocessing on the collected live video data, including removing stop words, word segmentation, and stemming operations, and converting text data into digital feature vectors; S202, feature selection: select corresponding feature representation methods, such as bag-of-words model and TF-IDF. These features describe the text information in the live video data; S203, labeling samples: label the collected live video data with the correct game type, such as Game 1 and Game 2. The specific game type depends on the specific situation. Use it as a training sample to construct a training set; S204, model training: use the training set to train the naive Bayes classifier model. During the training process, the model learns the probability distribution of the features of each category appearing in the data; S205, model application: use the trained model to classify the collected live video data. According to the feature vector of the text data, the model can predict the game type to which the video belongs;

[0047] Extended explanation: In addition to machine learning algorithms, you can also use rule matching to classify; for example, you can build a corresponding rule engine, such as judging the game type of the video based on keywords and game tags in the live broadcast room title. This method does not require a machine learning model, but requires setting rules; in addition, in order to improve the accuracy of classification, you can consider combining multiple features and algorithms for classification, such as combining barrage content, number of viewers and other information for comprehensive classification judgment, or use integrated learning algorithms, such as random forests and gradient boosting trees for classification; these methods can be selected and adjusted according to the needs of actual applications and data characteristics.

[0048] The evaluation and analysis module matches the corresponding live broadcast type according to the search content, obtains the required information in the live broadcast video data of the corresponding anchor under the live broadcast type, and extracts evaluation parameters from the required information of each corresponding type of anchor for a second time. The evaluation parameters include the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience. Based on the pre-processed evaluation parameters and the real-time number of violations and live broadcast duration of the corresponding type of anchor, a live broadcast data evaluation model is built to generate a live broadcast effect evaluation value Ir that reflects the popularity of the corresponding type of anchor;

[0049] Traditional methods usually only evaluate the live broadcast of the anchor based on basic indicators such as the number of viewers, while the scheme designed in this application comprehensively considers multiple evaluation parameters such as the number of viewers, the amount of barrage sent, the frequency of interaction between the anchor and the audience, etc., and can provide more comprehensive and multi-dimensional evaluation indicators. When obtaining the evaluation indicators, the real-time number of violations of the anchor is also considered, and the occurrence of violations is measured by the fluctuation of the number of viewers. This method makes the evaluation more real-time and flexible, and can observe the anchor's violations more carefully and monitor violations in a timely manner. It can also more accurately reflect the anchor's live broadcast effect and popularity through the evaluation indicators, namely the live broadcast effect evaluation index U.

[0050] Matching user search content with live broadcast type can be done in the following ways:

[0051] 1. Keyword matching: According to the keywords searched by the user, keyword matching is performed with the live broadcast content to find the relevant live broadcast types; for example, if the user searches for "fitness", it can match the fitness live broadcast type; 2. Content category matching: Matching is performed according to the classification of the live broadcast content, for example, if the user searches for "game", it can match the game live broadcast type; 3. Matching based on historical search records: According to the user's previous search records, historical search matching is performed to find the live broadcast type that the user may be interested in;

[0052] For example, a user searches for the keyword "game". Keyword matching: According to the keyword "game", the game live broadcast type can be matched; content classification matching: According to the classification of the game live broadcast content, the game live broadcast type can be matched; historical search record matching: If the user has previously searched for "game" related records, the game live broadcast type can be matched according to the historical search records; According to the above method, the user searches for the keyword "game" and can match all the anchors of the game live broadcast type. Usually, the system will sort them from large to small according to the corresponding live broadcast effect evaluation value Ir of each anchor;

[0053] It should be noted that the required information of the corresponding type of anchor of the live broadcast is obtained under the live broadcast type, and the required information here is: the live video data obtained in the information collection module, that is, the anchor information, the live broadcast room information, the number of viewers, the barrage content and the interactive information; the evaluation parameters are extracted twice from the required information of each corresponding type of anchor, and the extracted evaluation parameters are the current number of viewers in the number of viewers, the amount of barrage sent per unit time in the barrage content, and the frequency of interaction between the anchor and the audience in the interactive information;

[0054] The current number of viewers refers to the number of people who are watching the live broadcast of the corresponding anchor in real time. The number of barrages sent per unit time refers to the number of barrages sent per minute, for example: 5 / min. The frequency of interaction between the anchor and the audience is based on the number of prize questions initiated by the anchor. Therefore, the frequency of interaction between the anchor and the audience is the number of prize questions initiated by the anchor per hour, for example: 1 time / h.

[0055] Based on the preprocessed evaluation parameters and the real-time violation counts of the corresponding types of anchors, the preprocessing process of the evaluation parameters, the real-time violation counts, and the live broadcast duration is as follows: data cleaning and dimensionless processing of each data to remove the units of each parameter or data, so as to facilitate subsequent calculation and processing;

[0056] The process of obtaining the real-time violation count of the corresponding type of anchor is as follows: during the normal live broadcast process, the number of times the fluctuation value of the number of viewers of the live video at adjacent moments exceeds the preset standard threshold value F1, and the number of times the corresponding anchor has illegal content in the live broadcast room and has received warnings from the administrator F2. By adding F1 and F2, the sum obtained is the real-time violation count of the corresponding type of anchor; among which, the fluctuation value of the number of viewers of the live video at adjacent moments exceeds the preset standard threshold value, indicating that the number of viewers has suddenly dropped significantly, and the fluctuation value at adjacent moments = the number of online viewers at the previous moment - the number of online viewers at the next moment;

[0057] The process of generating the live broadcast effect evaluation value Ir is as follows:

[0058] S301. Generate a live broadcast effect evaluation index U based on the current number of viewers, the amount of barrage sent per unit time, the frequency of interaction between the anchor and the audience, the number of real-time violations of the corresponding type of anchor, and the live broadcast duration. The formula is as follows:

[0059]

[0060] In the formula (1), u Δ represents a positive index. gs, ds, and pc in formula (2) represent the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience, respectively. a1, a2, and a3 are the preset proportional coefficients of the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience, respectively. a1>a2>a3>0. The value ranges of a1, a2, and a3 are all 0 to 1. wc and zs in formula (3) represent the real-time violation times and live broadcast duration of the corresponding type of anchor, respectively. G is a constant correction coefficient, whose specific value can be adjusted and set by the user, or generated by the analysis function fitting, and the value range of G is 0 to 1.

[0061] S302: Obtaining the live broadcast effect evaluation index U at different times within T time. t , t represents the number of the live broadcast effect evaluation index at different times within time T, t = 1, 2, 3, 4, ..., n, n is a positive integer;

[0062] S303: Based on the average value of the live broadcast effect evaluation index at different times within the time T and the live broadcast effect evaluation index U t , calculate the live broadcast effect evaluation value Ir, the calculation formula is as follows:

[0063]

[0064] In the formula, It represents the average value of the live broadcast effect evaluation index at different times within time T.

[0065] The intelligent push module sorts the anchors corresponding to the same type of live broadcast data according to the size of the live broadcast effect evaluation index U, and generates a sorting table. The live broadcast effect evaluation index U in the sorting table is arranged from large to small. When the user searches for the anchor of this type, the corresponding anchors are pushed in sequence according to the sorting table;

[0066] Among them, the larger the live broadcast effect evaluation index U is, the higher the live broadcast effect and popularity of the live broadcast will be. Therefore, the system will give priority to pushing the anchor. The application scenario of this system is for game live broadcast. By processing the data generated during the live broadcast, different types of anchors in the game are evaluated, so that when users search for this type of anchor, they can push higher-quality anchors, which improves user viewing while increasing continuous traffic for the anchor and the platform, achieving a mutually beneficial effect, which reflects the intelligent design of the system.

[0067] The solution designed in this application is more comprehensive and accurate, and provides flexible and adjustable evaluation indicators. The intelligent push module can better meet the personalized needs of users. These advantages help to improve the user satisfaction and user stickiness of the live broadcast platform and improve the quality and diversity of live broadcast content.

[0068] Example 2: Please refer to Figure 2 Based on Example 1, this embodiment also provides a method for intelligently processing live video data, including the following specific steps:

[0069] S1. Use crawler technology to collect live video data from live data sources;

[0070] S2, automatically classifying the collected live video data by using machine learning algorithms;

[0071] S3. Match the corresponding live broadcast type according to the search content, and obtain the required information in the live broadcast video data of the corresponding anchor under the live broadcast type. Extract evaluation parameters from the required information of each corresponding type of anchor for the second time, and the evaluation parameters include the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience. According to the pre-processed evaluation parameters and the real-time number of violations and live broadcast duration of the corresponding type of anchor, build a live broadcast data evaluation model to generate the live broadcast effect evaluation value Ir of the corresponding anchor;

[0072] S4. Sort the anchors corresponding to the same type of live broadcast data according to the size of the live broadcast effect evaluation index U to generate a sorting table. The live broadcast effect evaluation index U in the sorting table is arranged from large to small. When searching for anchors of this type, push each corresponding anchor in turn according to the sorting table.

[0073] In the application, the several formulas involved are all calculated by removing the dimensions and taking their numerical values. The formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The formula is set by technical personnel in this field according to actual conditions.

[0074] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0075] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0076] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A live video data intelligent processing system, It is characterized in that include: The information collection module uses crawler technology to collect live video data from live data sources; The pre-classification module automatically classifies the collected live video data by using machine learning algorithms; The evaluation and analysis module matches the corresponding live broadcast type according to the search content, obtains the required information in the live broadcast video data of the corresponding anchor under the live broadcast type, and extracts evaluation parameters from the required information of each corresponding type of anchor for a second time. The evaluation parameters include the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience. Based on the pre-processed evaluation parameters and the real-time number of violations and live broadcast duration of the corresponding type of anchor, a live broadcast data evaluation model is built to generate the live broadcast effect evaluation value Ir of the corresponding anchor; The intelligent push module sorts the anchors corresponding to the same type of live broadcast data according to the size of the live broadcast effect evaluation index U, and generates a sorting table. The live broadcast effect evaluation index U in the sorting table is arranged from large to small. When searching for anchors of this type, the corresponding anchors are pushed in sequence according to the sorting table; The process of generating the live broadcast effect evaluation value Ir is as follows: S301. Generate a live broadcast effect evaluation index U based on the current number of viewers, the amount of barrage sent per unit time, the frequency of interaction between the anchor and the audience, the number of real-time violations of the corresponding type of anchor, and the live broadcast duration. The formula is as follows: In the formula (1), u Δ represents a positive index, gs, ds, and pc in formula (2) represent the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience, respectively, a1, a2, and a3 are the preset proportional coefficients of the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience, respectively, and a1>a2>a3>0, wc and zs in formula (3) represent the real-time violation times and live broadcast duration of the corresponding type of anchor, respectively, and G is a constant correction coefficient; S302: Obtaining the live broadcast effect evaluation index U at different times within T time. t , t represents the number of the live broadcast effect evaluation index at different times within time T, t = 1, 2, 3, 4, ..., n, n is a positive integer; S303: Based on the average value of the live broadcast effect evaluation index at different times within the time T and the live broadcast effect evaluation index U t , calculate the live broadcast effect evaluation value Ir, the calculation formula is as follows: In the formula, It represents the average value of the live broadcast effect evaluation index at different times within time T.

2. A live video data intelligent processing system according to claim 1, Features: The live broadcast data source includes various live broadcast platforms and websites, and the collected live broadcast video data includes the required information, including anchor information, live broadcast room information, number of viewers, barrage content and interactive information.

3. A live video data intelligent processing system according to claim 2, Features: The anchor information includes anchor ID, anchor nickname, and anchor level; the live broadcast room information includes live broadcast room ID, live broadcast room title, and live broadcast duration; the number of viewers includes the current number of viewers and the peak number of viewers; the barrage content includes the nickname of the sender of the barrage, the barrage sentence, the time when the barrage was sent, and the number of barrages sent per unit time; Interaction information includes the frequency of interaction between the host and the audience.

4. A live video data intelligent processing system according to claim 3, Features: The machine learning algorithms used in the pre-classification module include any one of a naive Bayes classifier, a support vector machine (SVM), and a convolutional neural network (CNN), and the pre-classification module is used to distinguish the game types to which the videos belong in the same type of game anchor video data.

5. A live video data intelligent processing system according to claim 4, Features: The process of preprocessing the evaluation parameters, the number of real-time violations, and the live broadcast duration is: data cleaning and dimensionless processing of each data.

6. A live video data intelligent processing system according to claim 5, Features: The process of obtaining the real-time number of violations of the corresponding type of anchor is as follows: during the normal live broadcast process, the number of times the fluctuation value of the number of viewers of the live video at adjacent moments exceeds the preset standard threshold F1, and the number of times the corresponding anchor has illegal content in the live broadcast room and received warnings from the administrator F2. By adding F1 and F2, the total obtained is the real-time number of violations of the corresponding type of anchor, among which the calculation method of the fluctuation value at adjacent moments is: the fluctuation value at adjacent moments = the number of online viewers at the previous moment - the number of online viewers at the next moment.

7. A method for intelligently processing live video data, using any system described in claims 1 to 6, Features: The method comprises the following steps: S1, collecting live video data from a live data source using crawler technology; S2, automatically classifying the collected live video data by using machine learning algorithms; S3, matching the corresponding live broadcast type according to the search content, and obtaining the required information in the live broadcast video data of the corresponding anchor under the live broadcast type, and extracting evaluation parameters from the required information of each anchor of the corresponding type for the second time; S4. Sort the anchors corresponding to the same type of live broadcast data according to the size of the live broadcast effect evaluation index U to generate a sorting table. The live broadcast effect evaluation index U in the sorting table is arranged from large to small. When searching for anchors of this type, push each corresponding anchor in turn according to the sorting table.

8. The method for intelligently processing live video data according to claim 7, It is characterized in that The evaluation parameters include the current number of viewers, the amount of barrage sent per unit time, and the frequency of interaction between the anchor and the audience. Based on the preprocessed evaluation parameters and the real-time number of violations and live broadcast duration of the corresponding type of anchor, a live broadcast data evaluation model is built to generate the live broadcast effect evaluation value Ir of the corresponding anchor.