Business product live stream data acquisition method and system

Abstract

The invention relates to the technical field of data acquisition, and particularly discloses a business product live broadcast stream data acquisition method and system, and the method comprises the steps: obtaining a plurality of groups of live broadcast source data in a business live broadcast scene, and enabling each group of live broadcast source data to comprise multimedia stream data, user operation data, transaction flow data and supply chain state data; acquiring differential data acquisition period information according to the live broadcast source data; acquiring a multi-batch data sample set according to the differential data acquisition period information; integrating and analyzing the multiple batches of data sample sets to obtain a data acquisition quality evaluation value; and acquiring a data acquisition judgment value according to the data acquisition quality evaluation value. By comprehensively collecting multi-source data such as multimedia stream, user operation, transaction flow and supply chain state, the market trend and user requirements can be accurately grasped, differential data collection is achieved, and collection efficiency and quality are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data acquisition, and in particular to a method and system for collecting live stream data of business products. Background Art

[0002] In today's digital business era, live streaming of business products, as a new and highly influential marketing model, is booming at an unprecedented speed. It breaks the limitations of traditional sales models in terms of time and space, and builds an efficient and intuitive communication bridge between enterprises and consumers. Consumers can watch product demonstrations in real time, obtain detailed information, and interact with the anchor. This new shopping experience greatly stimulates consumers' desire to purchase, making live streaming with goods an important means for many enterprises to increase sales and expand market share.

[0003] Existing live stream data acquisition technologies usually only focus on some data types, such as transaction flow data, ignoring key information such as multimedia stream data, user operation data, and supply chain status data. This makes it difficult for enterprises to grasp the overall picture of the live stream, and they cannot deeply understand user behavior, product display effects, and the real-time status of the supply chain. At the same time, the acquisition period lacks flexibility. Currently, a fixed acquisition period is mostly used, which cannot adapt to the change frequencies of different types of data and business requirements. For rapidly changing multimedia stream data and user operation data, fixed-period acquisition is likely to cause data lag and cannot timely reflect the real-time dynamics in the live stream. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for collecting live stream data of business products to solve the technical problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions: A method for collecting live stream data of business products, comprising: Obtaining multiple groups of live source data in a business live scene, wherein each group of the live source data includes multimedia stream data, user operation data, transaction flow data, and supply chain status data; Obtaining differential data acquisition period information according to the live source data; Obtaining multiple batches of data sample sets according to the differential data acquisition period information; Integrating and analyzing the multiple batches of data sample sets to obtain a data acquisition quality evaluation value; Obtaining a data acquisition judgment value according to the data acquisition quality evaluation value; Judging whether the data acquisition judgment value is greater than a preset threshold; If it is greater, storing the live source data in the core database; If it is less than, the live source data will be excluded.

[0006] Preferably, the step of obtaining multiple groups of live source data in the business live scene includes: Obtain multiple types of data access points in the business live scene; Obtain multiple groups of raw data according to multiple types of data access points; Obtain the format feature of each group of raw data, and judge the corresponding raw data as multimedia stream data according to the format feature; Obtain the user operation feature of each group of raw data, and judge the corresponding raw data as user operation data according to the user operation feature; Obtain the commodity transaction feature of each group of raw data, and judge the corresponding raw data as transaction flow data according to the commodity transaction feature; Obtain the commodity logistics feature of each group of raw data, and judge the corresponding raw data as supply chain status data according to the commodity logistics feature.

[0007] Preferably, the step of obtaining differential data collection period information according to the live source data includes: Obtain the data structure in the audio and video database according to the multimedia stream data; Obtain the timestamp feature information of key frames according to the data structure; Establish a key frame index and a scene label according to the timestamp feature information; Obtain the update frequency of the key frame index and the scene label, and use the update frequency as the first collection period; Obtain user operation record information according to the user operation data, wherein the user operation record information includes operation type information, operation time information, and operation object information; Obtain the association relationship between the operation type information, operation time information, and operation object information, and establish an operation-timestamp association index according to the association relationship; Obtain multiple activity levels according to the operation-timestamp association index, and obtain the second collection period of the user operation data according to the multiple activity levels.

[0008] Obtain order record information according to the transaction flow data, wherein the order record information includes order number information, commodity information, payment status information, and transaction time information; establish an order number-payment status service index according to the order record information; Obtain the service dependence degree of the service decision on the transaction data according to the order number-payment status service index, and obtain the third collection period according to the service dependence degree.

[0009] Obtain inventory - transportation record information based on the supply chain status data; Obtain the dynamic change information of the inventory - transportation record information; obtain the business - supply chain sensitivity according to the dynamic change information, and obtain the fourth collection period according to the business - supply chain sensitivity.

[0010] Summarize and integrate the first collection period, the second collection period, the third collection period, and the fourth collection period to generate differential data collection period information.

[0011] Preferably, the step of obtaining a multi - batch data sample set according to the differential data collection period information includes: Obtain multiple data collection tasks according to the differential data collection period information; Obtain collection configuration parameters according to each data collection task; Obtain the original traceable data sources corresponding to multimedia stream data, user operation data, transaction flow data, and supply chain status data according to the collection configuration parameters; Obtain a collection sequence according to each of the original traceable data sources, where the collection sequence includes a collection order, a collection frequency, and a collection interval; Perform high - frequency preliminary collection on each original traceable data source according to the preset collection sequence to obtain collection information; Integrate the collection information corresponding to multimedia stream data, user operation data, transaction flow data, and supply chain status data to form a multi - batch data sample set.

[0012] Preferably, the step of performing integrated analysis on the multi - batch data sample set to obtain a data collection quality evaluation value includes: Obtain the multi - batch data structure format and the multi - batch data content according to the multi - batch data sample set; Obtain format result information according to the multi - batch data structure format, where the format result information includes structured data result information, semi - structured data result information, and unstructured data result information; Obtain an information continuity evaluation value according to the structured data result information, the semi - structured data result information, and the unstructured data result information; Obtain content result information according to the multi - batch data content, where the content result information includes information integrity, information accuracy, information timeliness, information clarity, information relevance, and information continuity; Obtain the weight coefficients of information integrity, information accuracy, information timeliness, information clarity, information relevance, and information continuity in quality assessment according to the analytic hierarchy process, and obtain the content information evaluation value according to the weight coefficients of information integrity, information accuracy, information timeliness, information clarity, information relevance, and information continuity in quality assessment; Obtain the data collection quality evaluation value according to the information continuity evaluation value and the content information evaluation value.

[0013] Preferably, the step of obtaining the data collection judgment value according to the data collection quality evaluation value includes: Obtain the data proportions in the business product live broadcast service according to the multimedia stream data, user operation data, transaction flow data, and supply chain status data; Obtain the actual data volumes corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data; Obtain the data quality evaluation values corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data according to the data collection quality evaluation value, data proportion, and actual data volume; Obtain the total corrected data volume according to the multimedia stream data, user operation data, transaction flow data, and supply chain status data; Obtain the data collection judgment value according to the total corrected data volume and the data quality evaluation value.

[0014] The present invention also provides a business product live stream data collection system, including: A first acquisition module, configured to acquire multiple groups of live source data in a business live broadcast scenario, where each group of the live source data includes multimedia stream data, user operation data, transaction flow data, and supply chain status data; A second acquisition module, configured to acquire differential data collection cycle information according to the live source data; a third acquisition module, configured to acquire multiple batches of data sample sets according to the differential data collection cycle information; A fourth acquisition module, configured to perform integrated analysis on the multiple batches of data sample sets to obtain a data collection quality evaluation value; A fifth acquisition module, configured to obtain a data collection judgment value according to the data collection quality evaluation value; A judgment module, configured to judge whether the data collection judgment value is greater than a preset threshold; If it is greater, store the live source data in the core database; If it is less, eliminate the live source data.

[0015] Preferably, the first acquisition module includes: The first acquisition unit is configured to acquire multiple types of data access points in the business live broadcast scenario; The second acquisition unit is configured to acquire multiple groups of raw data according to the multiple types of data access points; The first determination unit is configured to acquire the format features of each group of raw data, and determine the corresponding raw data as multimedia stream data according to the format features; The second determination unit is configured to acquire the user operation features of each group of raw data, and determine the corresponding raw data as user operation data according to the user operation features; The third determination unit is configured to acquire the commodity trading features of each group of raw data, and determine the corresponding raw data as transaction flow data according to the commodity trading features; The fourth determination unit is configured to acquire the commodity logistics features of each group of raw data, and determine the corresponding raw data as supply chain status data according to the commodity logistics features.

[0016] Preferably, the second acquisition module includes: The third acquisition unit is configured to acquire the data structure in the audio and video database according to the multimedia stream data; The fourth acquisition unit is configured to acquire the timestamp feature information of the key frames according to the data structure; The establishment unit is configured to establish a key frame index and a scene label according to the timestamp feature information; The fifth acquisition unit is configured to acquire the update frequency of the key frame index and the scene label, and use the update frequency as the first acquisition period; The sixth acquisition unit is configured to acquire multiple activity levels according to the user operation data, and acquire the second acquisition period of the user operation data according to the multiple activity levels; The seventh acquisition unit is configured to acquire the business dependence degree of the business decision on the transaction data according to the transaction flow data, and the third acquisition period according to the business dependence degree; The eighth acquisition unit is configured to acquire the business-supply chain sensitivity according to the supply chain status data, and acquire the fourth acquisition period according to the business-supply chain sensitivity; The first generation unit is configured to summarize and integrate the first acquisition period, the second acquisition period, the third acquisition period, and the fourth acquisition period to generate differential data acquisition period information.

[0017] Preferably, the third acquisition module includes: The ninth acquisition unit is configured to acquire multiple data acquisition tasks according to the differential data acquisition period information; The first acquisition unit is configured to acquire acquisition configuration parameters according to each data acquisition task; A tenth acquisition unit, configured to obtain original traceability data sources corresponding to multimedia stream data, user operation data, transaction flow data, and supply chain status data according to the acquisition configuration parameters; A traceability unit, configured to obtain an acquisition sequence according to each of the original traceability data sources, where the acquisition sequence includes an acquisition order, an acquisition frequency, and an acquisition interval; A second acquisition unit, configured to perform high-frequency preliminary acquisition on each original traceability data source according to a preset acquisition sequence to obtain acquisition information; A third acquisition unit, configured to integrate the acquisition information corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data to form a multi-batch data sample set.

[0018] The beneficial effects of this application are as follows: By comprehensively collecting multi-source data such as multimedia streams, user operations, transaction flows, and supply chain statuses, the present invention can accurately grasp market trends and user needs. Secondly, by means of establishing indexes, determining acquisition cycles, etc., differential data acquisition is realized, the acquisition efficiency and quality are improved, resource waste is avoided, and the acquired data is ensured to meet business requirements. Furthermore, the data quality is evaluated from multiple dimensions, and the data is screened according to the evaluation results to ensure the high quality of the data in the core database and provide a reliable basis for decision-making. Finally, the entire process and system cooperate with each other to optimize live broadcast operations, help enterprises adjust strategies, improve user experience, and increase sales conversion rates. Description of the Drawings

[0019] Figure 1 It is a schematic flowchart of the method according to an embodiment of this application.

[0020] Figure 2 It is a schematic structural diagram of the system according to an embodiment of this application.

[0021] The implementation, functional features, and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Embodiments

[0022] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0023] As Figure 1 shown, this application provides a method for collecting live stream data of business products, which is applied to a circuit component feature database and includes: S1. Obtain multiple groups of live stream source data in a business live broadcast scenario, where each group of the live stream source data includes multimedia stream data, user operation data, transaction flow data, and supply chain status data; S2. Obtain differential data acquisition cycle information according to the live stream source data; S3, acquiring multiple batches of data sample sets according to the differentiated data collection cycle information; S4, integrating and analyzing the multiple batches of data sample sets to obtain a data collection quality assessment value; S5. Obtaining a data collection judgment value according to the data collection quality evaluation value; S6, determining whether the data collection judgment value is greater than a preset threshold; If it is greater, the live source data will be stored in the core database; If it is less than that, the live source data will be removed.

[0024] As described in the above steps S1 - S6, the present invention obtains multiple groups of live source data in the business live broadcast scenario. In actual operation, first find numerous data access points in the business live broadcast scenario, such as live platform interfaces, sensor interfaces, etc. Then, collect multiple groups of raw data from these access points and classify them according to different characteristics. For example, data in video and audio formats is determined to be multimedia stream data, such as the pictures and commentary voices when the host shows products during the live broadcast; data generated by users' operations such as clicks, comments, and orders belongs to user operation data; data related to commodity transactions such as order amounts and commodity quantities is transaction flow data; and information related to commodity logistics such as inventory quantities and logistics locations is classified as supply chain status data. By comprehensively collecting this data, it provides rich materials for subsequent analysis, solves the problems of single data and incomplete information in traditional collection methods, and makes our understanding of the live broadcast more complete. For example, in a live broadcast of electronic products, we can understand the product display effect, user participation, sales situation, and the support ability of the supply chain through this data. Then, obtain differential data collection cycle information according to the live source data. For multimedia stream data, first clarify its data structure in the audio - video database, obtain the timestamp feature information of key frames from it, establish a key frame index and a scene label, and take the update frequency of the key frame index and the scene label as the first collection cycle. For user operation data, collect user operation record information, analyze the correlation between operation types, times, and objects, construct an operation - timestamp correlation index, and determine the second collection cycle according to the activity level. Transaction flow data establishes an order number - payment status business index based on order record information and determines the third collection cycle according to the dependence degree of business decisions on transaction data. Supply chain status data determines the fourth collection cycle by obtaining the dynamic changes of inventory - transportation record information and according to the business - supply chain sensitivity. Finally, summarize and integrate these four collection cycles to form differential data collection cycle information. This step solves the problem that the traditional collection cycle is fixed and cannot adapt to different data change frequencies and business requirements, improves the data collection efficiency and quality, and avoids resource waste. For example, in a beauty live broadcast, a reasonable collection cycle can be formulated according to the characteristics of different data to ensure the timeliness and accuracy of the data. After that, obtain multiple batches of data sample sets according to the differential data collection cycle information. Generate multiple data collection tasks according to the previously determined differential data collection cycle information, and configure parameters such as collection time and scope for each task. Find the corresponding original traceable data sources according to these parameters, and determine the collection sequence, number of times, and intervals, etc. for each data source. Then, conduct high - frequency preliminary collection on each data source according to the preset collection sequence, and finally integrate the collection information of different types of data to form multiple batches of data sample sets. This step ensures the timeliness and accuracy of the data, provides reliable data for subsequent analysis, and solves the problems of untimely and inaccurate data collection and single samples that cannot reflect the real situation.For example, in food live streaming, collecting multiple batches of data samples according to a reasonable collection cycle can comprehensively reflect the product display effect, user purchase behavior, and real-time status of the supply chain during the live streaming process. Then, integrating and analyzing the multiple batches of data sample sets to obtain a data collection quality evaluation value. Obtain the data structure format and content from the multiple batches of data sample sets, and based on the data structure format, obtain structured, semi-structured, and unstructured data result information to evaluate the information continuity. Obtain content result information such as information integrity, accuracy, timeliness, clarity, relevance, and continuity from the data content. Use the analytic hierarchy process to determine the weight coefficients of each dimension in the quality evaluation, calculate the content information evaluation value, and finally combine the information continuity evaluation value and the content information evaluation value to obtain the data collection quality evaluation value. This step can comprehensively evaluate the data collection quality, help discover problems in the collection process, and provide a basis for optimizing the collection strategy. For example, in a home furnishing products live streaming, through the analysis of multiple batches of data sample sets, problems in some data are found, and then the advantages and disadvantages of data collection are clarified. Subsequently, obtain the data collection judgment value according to the data collection quality evaluation value. First, determine the data proportions of multimedia stream data, user operation data, transaction flow data, and supply chain status data in the business product live streaming service, and then obtain the actual data volumes of various types of data. Combine the data collection quality evaluation value, data proportion, and actual data volume to calculate the data quality evaluation values of various types of data. Then, obtain the total corrected data volume corresponding to each type of data, and obtain the data collection judgment value based on the total corrected data volume and the data quality evaluation value. This step provides a quantitative basis for data storage and processing, helps screen out high-quality data, improves the data quality in the database, and ensures the accuracy of subsequent analysis and decision-making. For example, in an automotive products live streaming, calculate the data collection judgment value through these data to provide decision support for data storage and processing. Finally, determine whether the data collection judgment value is greater than a preset threshold. Compare the calculated data collection judgment value with the preset threshold. If it is greater than the threshold, it means that the data quality meets the requirements, and the live streaming source data is stored in the core database; if it is less than the threshold, it indicates that there are problems with the data quality, and the live streaming source data is excluded. This step realizes the effective screening and management of data, ensures the data quality in the core database, provides reliable data support for subsequent business applications, and avoids the negative impact of low-quality data on data analysis and decision-making. For example, in a clothing live streaming, decide whether to keep or discard the data according to the judgment result to ensure that the data stored in the database can accurately reflect the real situation of the live streaming and provide strong support for business decisions.

[0025] In one embodiment, the step of obtaining multiple groups of live streaming source data in the business live streaming scenario includes: S101. Obtain multiple types of data access points in the business live streaming scenario; S102. Obtain multiple groups of original data according to the multiple types of data access points; S103. Obtain the format features of each group of original data, and determine the corresponding original data as multimedia stream data according to the format features; S104. Obtain the user operation features of each group of original data, and determine the corresponding original data as user operation data according to the user operation features; S105. Obtain the commodity transaction features of each group of original data, and determine the corresponding original data as transaction flow data according to the commodity transaction features; S106. Obtain the commodity logistics features of each group of original data, and determine the corresponding original data as supply chain status data according to the commodity logistics features.

[0026] As described in the above steps S101 - S106, the present invention obtains multiple types of data access points in the business live - streaming scenario. It conducts in - depth research on the architecture and business processes of the business live - streaming system, comprehensively sorts out all sources of data generated during the live - streaming process, including internal interfaces of the live - streaming platform, third - party service interfaces, and interactive device interfaces in the live - streaming room, etc. By determining these data access points, it ensures that no important data sources are missed. For example, in a beauty live - streaming, the live - streaming platform interface can obtain the live - streaming video and sound, the payment platform interface can obtain transaction data, the logistics platform interface can get product distribution information, and the interactive device interface in the live - streaming room can collect operation data such as user comments and likes. This step solves the problem of a single data - collection entry and lays a foundation for comprehensive data collection. Next, multiple groups of raw data are obtained according to the multiple types of data access points. For the determined various data access points, appropriate data - collection programs and transmission protocols are used. For example, data is obtained from the live - streaming platform interface through the HTTP / HTTPS protocol, and data is obtained from the third - party platform using the API interface, etc. Data is collected in real - time or at regular intervals from each access point, and preliminary verification and sorting are carried out during the collection process. For example, in a live - streaming of electronic products, different - time - period live - streaming video segments and audience interaction records are obtained from the live - streaming platform interface, detailed information of each order is obtained from the payment platform interface, and product inventory and transportation track data are obtained from the logistics platform interface. This step ensures the richness and diversity of data and solves the problem of one - sided analysis caused by insufficient data acquisition. Then, the format characteristics of each group of raw data are obtained, and the corresponding raw data is judged as multimedia stream data according to the format characteristics. Data - format recognition technologies are used, such as analyzing file extensions, data - header information, etc., to determine whether the raw data conforms to common multimedia formats, such as MP4 for video and MP3 for audio. Then, further check whether the data content contains live - streaming video and sound information. If it meets the criteria, it is determined as multimedia stream data. For example, in a live - streaming of household items, a group of collected data is determined to be in the MP4 format after analyzing the file extension and data - header information, and the content is the picture of the host showing the product and the explanation sound, so it is determined as multimedia stream data. This helps to focus on the display content of the live - streaming and evaluate the audio - visual effect of the live - streaming. After that, the user - operation characteristics of each group of raw data are obtained, and the corresponding raw data is judged as user - operation data according to the user - operation characteristics. The raw data is deeply analyzed to extract feature information related to user operations, such as click events, comment content, sharing behaviors, following actions, etc. By identifying and analyzing these features, it is determined whether the raw data belongs to user - operation data. For example, in a live - streaming of clothing, if the raw data contains information such as users clicking on the purchase link, asking about size and color in the comment area, and sharing the live - streaming link, this data will be determined as user - operation data.This step can deeply understand the user's behavior patterns and interest preferences during the live broadcast, providing a basis for optimizing the live broadcast interaction session and product recommendation strategy. After that, obtain the commodity transaction characteristics of each group of original data, and determine the corresponding original data as transaction flow data according to the commodity transaction characteristics. Extract characteristic information related to commodity transactions such as order number, commodity name, transaction amount, purchase quantity, payment method, and transaction time from the original data. By comprehensively judging these characteristics, determine whether the original data belongs to transaction flow data. For example, in a food live broadcast, if the original data contains complete information such as the order number, purchase quantity, payment amount, payment time, and payment method of a certain user purchasing a specific food, it can be determined as transaction flow data. This helps to accurately grasp the sales situation of the live broadcast, evaluate the commercial value of the live broadcast, and provide an important basis for the enterprise to formulate relevant strategies. Finally, obtain the commodity logistics characteristics of each group of original data, and determine the corresponding original data as supply chain status data according to the commodity logistics characteristics. Extract characteristic information related to commodity logistics such as inventory quantity, inventory location, shipping time, transportation method, logistics order number, and logistics track from the original data. By analyzing and judging these characteristics, determine whether the original data belongs to supply chain status data. For example, in a digital product live broadcast, if the original data contains information such as the inventory quantity, storage warehouse location, shipping time, and logistics order number of a certain digital product, these data will be determined as supply chain status data. This step can monitor the supply chain link of the product in real time, ensure the stability of product supply, and timely discover and solve problems in the supply chain.

[0027] In one embodiment, the step of obtaining the differential data collection period information according to the live broadcast source data includes: S201. Obtain the data structure in the audio and video database according to the multimedia stream data; S202. Obtain the timestamp feature information of the key frames according to the data structure; S203. Establish a key frame index and a scene label according to the timestamp feature information; S204. Obtain the update frequency of the key frame index and the scene label, and use the update frequency as the first collection period; S205. Obtain the user operation record information according to the user operation data, where the user operation record information includes operation type information, operation time information, and operation object information; S206. Obtain the association relationship between the operation type information, the operation time information, and the operation object information, and establish an operation-timestamp association index according to the association relationship; S207. Obtain multiple activity levels according to the operation-timestamp association index, and obtain the second collection period of the user operation data according to the multiple activity levels; S208. Obtain order record information according to the transaction flow data, where the order record information includes order number information, commodity information, payment status information, and transaction time information; establish an order number - payment status service index according to the order record information; S209. Obtain the service dependence degree of the service decision on the transaction data according to the order number - payment status service index, and according to the third collection period of the service dependence degree; S210. Obtain inventory - transportation record information according to the supply chain status data; S211. Obtain the dynamic change information of the inventory - transportation record information; obtain the service - supply chain sensitivity according to the dynamic change information, and obtain the fourth collection period according to the service - supply chain sensitivity; S212. Summarize and integrate the first collection period, the second collection period, the third collection period, and the fourth collection period to generate differential data collection period information.

[0028] As described in the above steps S201 - S212, the present invention obtains the data structure in the audio - video database according to the multimedia stream data; by querying relevant information of the audio - video database, such as viewing the database design document or using management tools, to clarify how the multimedia stream data is stored in the database, including table structure, field settings, and the relationships between them, etc. Just like in a live broadcast of an electronic product launch event, through this way, it can be known which fields store video data, audio data, and relevant timestamps and other information respectively, laying a foundation for subsequent operations. This step solves the problem of difficult extraction of key data due to the lack of understanding of the data storage structure. Only by being clear about the data structure can multimedia stream data be processed more efficiently. Then, according to the data structure, obtain the timestamp feature information of the key frames. After clarifying the storage structure of the multimedia stream data, find the fields storing the key frame timestamp information, and then extract these timestamp information through database query statements or specialized data processing tools. For example, in a beauty live broadcast, use the query statement to extract the timestamps of key scenes such as the host trying out products from the corresponding fields, facilitating quick positioning of these important segments in the future. This step solves the problem of difficultly finding the key scene time points quickly in a large amount of multimedia data, providing convenience for in - depth analysis of the live broadcast content. Then, establish key frame indexes and scene labels according to the timestamp feature information. Based on the obtained key frame timestamps, create indexes in the database or data storage system, and at the same time add scene labels to the key frames in combination with the live broadcast content. For example, in a car live broadcast, a certain key frame shows the acceleration performance of the car at a specific timestamp. Establish an index based on this timestamp and add a scene label of "car acceleration performance display". In this way, whether users search for specific scenes or operators analyze live broadcast data, it becomes more convenient and efficient, solving the problem of difficult retrieval and management of multimedia data. After that, obtain the update frequency of the key frame indexes and scene labels, and use the update frequency as the first acquisition period. Regularly check the update situation of the key frame indexes and scene labels, count the number of updates within a period of time, divide the number of updates by this period of time to get the update frequency, and this frequency is the acquisition period of the multimedia stream data. For example, in a live broadcast of a sports event, key scenes frequently appear during the game. After statistics, it is found that there is an update on average every 3 minutes. Then the acquisition period of the multimedia stream data is set to 3 minutes. This step solves the problem of unreasonable acquisition period of multimedia stream data, dynamically adjusts the acquisition period according to the changes in the live broadcast content, and avoids resource waste. Then, obtain user operation record information according to user operation data. Extract the records containing information such as operation type, operation time, and operation object from the database table or log file storing user operation data.For example, in an e-commerce live stream, by querying the database, it is possible to obtain operation records such as when users liked, commented on, or placed orders for which products, providing detailed data support for analyzing user behavior and solving the problem of insufficient understanding of users' live stream behavior. Then, obtain the correlation relationships among the operation type information, operation time information, and operation object information, and establish an operation-timestamp correlation index based on the correlation relationships. Analyze the internal connections among these information, taking the operation time as the main clue, associate different operation types and operation objects at the same time point, and then create an index in the database according to this correlation rule. For example, in an online education live stream, based on the operation records that students clicked on the courseware link and asked questions at a certain time point, establish an operation-timestamp correlation index to facilitate subsequent querying and analysis of students' learning behavior and solve the problem of difficult querying and analysis of user operation data. Next, obtain multiple activity levels according to the operation-timestamp correlation index, and obtain the second collection period of user operation data based on the multiple activity levels. Using the established correlation index, count the number of user operations in different time periods. The time period with a large number of operations indicates high user activity, while the time period with a small number of operations indicates low activity. Set different collection periods according to the level of activity. The collection period is short for high-activity time periods and long for low-activity time periods. For example, in a game live stream, the audience operates frequently in the first 30 minutes after the live stream starts, which is a high-activity period, and the collection period is set to 5 minutes; the operation decreases during a certain period in the middle, which is a low-activity period, and the collection period is set to 30 minutes. In this way, collect resources reasonably according to the user activity level. After that, obtain order record information according to the transaction flow data, and establish an order number-payment status business index according to the order record information. Extract information such as order number, product information, payment status, and transaction time from the database table storing the transaction flow data, and then create a business index in the database according to the correlation relationship between the order number and the payment status. For example, in an e-commerce promotion live stream, after establishing such an index, financial personnel can quickly count the number of paid orders, and the inventory management department can also timely understand which orders have been paid to arrange shipments, solving the problem of inconvenient querying and management of transaction flow data. Then, obtain the business dependence degree of business decisions on transaction data according to the order number-payment status business index, and determine the third collection period according to the business dependence degree. Analyze the demand situations of different business departments in the enterprise for transaction data. For example, the sales department needs to understand the order situation in real time to adjust strategies and has a high dependence degree; the market department has a low dependence degree when conducting long-term analysis. According to this difference in dependence degree, set different collection periods for different departments. The collection period for the sales department is short, and the collection period for the market department is long. For example, in a clothing e-commerce enterprise, the collection period for the sales department is set to 15 minutes, and the collection period for the market department is set to once a week, optimizing resource allocation while meeting the needs of different departments. Then, obtain inventory-transportation record information according to the supply chain status data.Extract information such as the inventory quantity, storage location, shipping order number, shipping route, and current location of the product from the database or system storing the supply chain status data. In a live e-commerce furniture broadcast, the enterprise learned about the inventory and shipping status of a certain sofa in this way, providing a basis for decision-making and solving the problem of the enterprise's untimely and inaccurate understanding of the supply chain status. Next, obtain the dynamic change information of the inventory-shipping record information; obtain the business-supply chain sensitivity based on the dynamic change information, and obtain the fourth collection cycle based on the business-supply chain sensitivity. Continuously monitor the changes in inventory and shipping information, such as increases or decreases in inventory quantity and changes in shipping locations, and evaluate the sensitivity of the business to supply chain changes based on the frequency and magnitude of the changes. For fresh products, inventory and shipping changes have a large impact and high sensitivity, and the collection cycle is set to every hour; for non-fresh products, the sensitivity is low, and the collection cycle is set to every day. This can accurately monitor the supply chain status, detect problems in a timely manner and take measures. Finally, summarize and integrate the first collection cycle, the second collection cycle, the third collection cycle, and the fourth collection cycle to generate differentiated data collection cycle information. Combine the collection cycles determined for multimedia stream data, user operation data, transaction flow data, and supply chain status data to form a complete set of differentiated data collection cycle solutions that meet the characteristics of different data and business requirements. This solution can optimize the data collection process, improve the effectiveness and value of the data, and provide strong support for subsequent analysis and decision-making in business live broadcasts.

[0029] In one embodiment, the step of obtaining a multi-batch data sample set according to the differentiated data collection cycle information includes: S301. Obtain multiple data collection tasks according to the differentiated data collection cycle information; S302. Obtain collection configuration parameters according to each data collection task; S303. Obtain the original traceable data sources corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data according to the collection configuration parameters; S304. Obtain a collection sequence according to each of the original traceable data sources, where the collection sequence includes a collection order, a collection frequency, and a collection interval; S305. Perform high-frequency preliminary collection on each original traceable data source according to the preset collection sequence to obtain collection information; S306. Integrate the collection information corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data to form a multi-batch data sample set.

[0030] As described in the above steps S301 - S306, the present invention obtains multiple data collection tasks according to the differential data collection cycle information. By analyzing the existing differential data collection cycle information in detail, it is transformed into specific data collection tasks. For example, it is known that multimedia stream data is collected every 3 minutes, user operation data is collected every 15 seconds during the active live period and every 5 minutes during the inactive period, transaction flow data is collected in real - time, and for supply chain status data, inventory data is collected daily and logistics data is collected every 3 hours. Based on these cycle information, collection tasks are set for different types of data respectively. For example, a task of starting collection every 3 minutes is set for multimedia stream data, corresponding frequency collection tasks are set for user operation data during different active periods, a continuous real - time collection task is set for transaction flow data, and timed collection tasks are set for inventory and logistics data of supply chain status data respectively. In this way, the collection tasks are closely matched with the data characteristics and business requirements, avoiding blind collection, improving the pertinence and efficiency of the collection work, and solving the problem that the collection tasks are out of touch with the actual needs. Then, collection configuration parameters are obtained according to each data collection task. For each collection task determined in the previous step, the collection configuration parameters are clarified by combining the characteristics of the data and the business objectives. Taking the multimedia stream data collection task as an example, considering the need to clearly display product details and ensure the viewing experience, the collection resolution is set to 1080p, the frame rate is set to 30fps, and the audio sampling rate is set to 44.1kHz; for the user operation data collection task, to ensure comprehensive recording of user behavior, the set event types for collection include clicks, comments, placing orders, etc., and the collection devices cover the PC side and the mobile side; for the transaction flow data collection task, fields such as order number, product name, transaction amount, payment method, and transaction time are determined for collection; for the supply chain status data collection task, according to the data source and update frequency, it is set that inventory data is collected from a specific interface of the inventory management system, logistics data is collected from the logistics system interface, and the request frequency is clarified. The determination of these parameters realizes the refined control of the data collection process, ensures the quality of the collected data, and solves the problem of lack of precise control in the collection process. Then, according to the collection configuration parameters, the original traceable data sources corresponding to multimedia stream data, user operation data, transaction flow data, and supply chain status data are obtained. Based on the collection configuration parameters determined previously, the original traceable data sources are found by combining the data type and the business system architecture. The original traceable data source of multimedia stream data is usually the video stream and audio stream servers of the live platform; user operation data comes from the user behavior log system of the live platform; transaction flow data originates from the transaction database of the e - commerce platform; for supply chain status data, inventory data comes from the enterprise internal inventory management system, and logistics data comes from the logistics transportation system database.When determining these data sources, factors such as data access rights and interface stability also need to be considered to ensure the reliability and stability of data acquisition, solving the problems of unclear data sources and difficult-to-guarantee reliability. After that, according to each original traceable data source, a collection sequence is obtained. The collection sequence includes the collection order, the number of collections, and the collection interval. For different original traceable data sources, the collection sequence is determined based on their characteristics and the requirements of the collection task. Due to the large amount of data and strong real-time nature of multimedia stream data, it is set to collect video frames and audio segments in sequence according to the time order. The number of collections is determined according to the collection cycle (for example, a 3-minute collection cycle, collecting data within 3 minutes each time), and the collection interval is continuous collection; for user operation data, due to random behavior, real-time capture of operation events is set, and the number of collections is determined according to the user operation frequency (more frequent collection during active periods and less collection during inactive periods), and the collection interval is adjusted according to the activity level; for transaction flow data, because it is important and real-time, each transaction record is obtained in real-time, and it is collected every time there is a new transaction, and the collection interval is 0; in the supply chain status data, inventory data is collected once a day at a fixed time, and the collection interval is 24 hours. Logistics data is collected according to changes in transportation nodes, and the number of collections is determined according to the number of nodes, and the collection interval is 3 hours. A reasonable collection sequence makes the collection process orderly and efficient, ensuring data quality and solving the problems of chaotic collection process and unstable data quality. After that, high-frequency preliminary collection is performed on each original traceable data source according to the preset collection sequence to obtain collection information. Use appropriate data collection tools or programs to perform high-frequency collection on the original traceable data source according to the established collection sequence. Multimedia stream data is collected sequentially from the video stream server of the live platform through web crawler technology; write a script to monitor the user behavior log system in real-time to obtain user operation data; use a database connection tool to query the transaction database in real-time to obtain transaction flow data; use data interface call technology to collect supply chain status data from the inventory management system and the logistics transportation system. During the collection process, preliminary verification and collation of the data are carried out to ensure the integrity and accuracy of the data. Such high-frequency collection can capture data changes in a timely manner, providing sufficient materials for subsequent analysis and solving the problems of untimely data collection and inaccurate capture of changes. Finally, the collection information corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data is integrated to form a multi-batch data sample set. First, clean and preprocess the collected various types of data. Denoise and clip the multimedia stream data, remove duplicates and classify the user operation data, unify the format and verify the transaction flow data, and process missing values and standardize the supply chain status data. Then, based on time, associate and integrate different types of data within the same time period. For example, integrate the live broadcast screen at a certain moment, the user operation data before and after that moment, the corresponding transaction flow data, and the supply chain status data into a data sample, and repeat this process continuously to form a multi-batch data sample set.These sample sets comprehensively reflect the overall situation of the live broadcast, helping to explore data correlations, providing a strong basis for enterprise decision-making, and solving the problem of scattered data that cannot be comprehensively analyzed.

[0031] In one embodiment, the step of integrating and analyzing the multi-batch data sample sets to obtain a data collection quality evaluation value includes: S401. Obtain multi-batch data structure formats and multi-batch data contents according to the multi-batch data sample sets; S402. Obtain format result information according to the multi-batch data structure formats, where the format result information includes structured data result information, semi-structured data result information, and unstructured data result information; S403. Obtain an information continuity evaluation value according to the structured data result information, semi-structured data result information, and unstructured data result information; S404. Obtain content result information according to the multi-batch data contents, where the content result information includes information integrity, information accuracy, information timeliness, information clarity, information relevance, and information continuity; S405. Obtain the weight coefficients of information integrity, information accuracy, information timeliness, information clarity, information relevance, and information continuity in quality evaluation according to the analytic hierarchy process, and obtain a content information evaluation value according to the weight coefficients of information integrity, information accuracy, information timeliness, information clarity, information relevance, and information continuity in quality evaluation; S406. Obtain a data collection quality evaluation value according to the information continuity evaluation value and the content information evaluation value.

[0032] As described in the above steps S401 - S406, the present invention obtains multi - batch data structure formats and multi - batch data contents from multi - batch data sample sets. With the help of data parsing techniques and tools, the data sample sets are deeply analyzed. For the data stored in the database, by viewing the database metadata, the field definitions, data types, and their relationships in the data table are clarified. For example, view the relevant information of fields such as product model and price in the product sales data table; for the data in the form of files, judge its format according to the file extension and internal identifier. For example, judge whether the user comment data is in CSV format or JSON format, etc. At the same time, directly read the data content and sort and classify it, laying a solid foundation for subsequent analysis. Just like in a data sample set of a mobile phone live broadcast, structured data of product parameters, semi - structured data of user evaluations, and unstructured data of live broadcast videos can be obtained, as well as their respective detailed contents, solving the problem of being unable to effectively analyze the quality due to lack of understanding of the data structure and content. Then, according to the multi - batch data structure formats, format result information is obtained, which includes structured, semi - structured, and unstructured data result information. For different types of structured data, their respective adapted parsing methods are used. For structured data, according to its clear fields and data types, data pattern information is extracted. For example, obtain the number of fields, association relationships, etc. from the sales data table; for semi - structured data, use its tags or format conventions to parse out the hierarchical structure and key information. For example, parse JSON - formatted user comment data to obtain comment content, commenter ID, etc.; for unstructured data, with the help of specific technical means, such as using natural language processing technology to extract text keywords and using image processing technology to extract image features. For example, in the data of a cosmetics live broadcast, unique result information is obtained from the structured product information data, semi - structured user comment data, and unstructured live broadcast video data respectively, which helps to deeply understand the characteristics and quality status of different types of data, solving the problem of being difficult to deeply evaluate the quality by generally analyzing the data structure format. Then, according to the structured, semi - structured, and unstructured data result information, an information continuity evaluation value is obtained. For structured data, check its coherence in time series or logical association. For example, check whether the records of sales data at different time points are complete; for semi - structured data, check the coherence of relevant information between different records. For example, whether the logic of different users' comments on the same product is coherent; for unstructured data, analyze the coherence of video frames or text content. For example, whether there are dropped frames in the live broadcast video and whether the semantics of user comments are smooth. Combining these aspects, an evaluation score is given for the information continuity of the data.In a clothing live stream, if there are missing sales data, chaotic user comment logic, or lagging live stream videos, it will affect the information continuity evaluation value. This step solves the problem that discontinuous data affects the accuracy of the analysis results. After that, content result information is obtained based on multiple batches of data content, covering information integrity, accuracy, timeliness, clarity, relevance, and continuity. Check whether the data contains necessary information to evaluate integrity, such as whether key product attributes are missing in the sales data; compare the data with reliable data sources to judge accuracy, like verifying product price data; evaluate timeliness based on the data generation time and usage scenario requirements, for example, determining whether the live sales data can reflect the current situation; evaluate clarity from aspects such as text comprehensibility and image recognizability, such as checking whether the live stream screen is clear; analyze the degree of association between the data and the business objective to evaluate relevance, such as calculating the proportion of content related to product functions in user comments; evaluate continuity from the perspective of content logic and check whether user comments are coherent. In the data evaluation of a home furnishing product live stream, if there are situations such as missing data, price deviations, and update delays, it will affect the evaluation results of each dimension, solving the problem that single-dimensional evaluation cannot comprehensively measure the quality of data content. After that, according to the analytic hierarchy process, the weight coefficients of information integrity, accuracy, timeliness, clarity, relevance, and continuity in quality evaluation are obtained, and based on this, the content information evaluation value is obtained. First, construct a hierarchical structure model including the target layer, criterion layer, and scheme layer. Compare the relative importance of each index in the criterion layer through expert scoring or data analysis, construct a judgment matrix and conduct a consistency test to ensure reasonable judgment, and then calculate the weight coefficients of each index. Finally, add the evaluation values of each index multiplied by their corresponding weight coefficients to obtain the content information evaluation value. For example, in the data evaluation of a digital product live stream, after determining the weights of each index, calculate the content information evaluation value by combining the evaluation values of each index before, solving the problem of unreasonable weight determination in multi-index evaluation. Finally, obtain the data collection quality evaluation value based on the information continuity evaluation value and the content information evaluation value. According to the data characteristics and business requirements, determine the fusion method of the two, such as using the weighted average method. Set weights for the information continuity evaluation value and the content information evaluation value respectively, multiply them by their weights and then add them to obtain the data collection quality evaluation value. In the data evaluation of a food live stream, assuming the information continuity evaluation value is 7 points and the content information evaluation value is 8 points, calculate the data collection quality evaluation value according to the set weights, comprehensively reflecting the quality level of data collection and solving the problem that separate evaluation cannot comprehensively measure the quality of data collection.

[0033] In one embodiment, the step of obtaining a data collection judgment value according to the data collection quality evaluation value includes: S501. Obtain the data proportion in the business product live stream service according to the multimedia stream data, user operation data, transaction flow data, and supply chain status data; S502. Obtain the actual data volumes corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data; S503. Obtain the data quality evaluation values corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data according to the data acquisition quality evaluation value, data ratio, and actual data volume; S504. Obtain the total corrected data volume according to the multimedia stream data, user operation data, transaction flow data, and supply chain status data; S505. Obtain the data acquisition judgment value according to the total corrected data volume and the data quality evaluation value.

[0034] As described in the above steps S501 - S505, the present invention obtains the data proportion in the business product live broadcast service based on multimedia stream data, user operation data, transaction flow data, and supply chain status data. This requires a comprehensive review of the business product live broadcast service process and an in - depth analysis of the roles played by each type of data in promoting business development, influencing user experience, determining sales results, and ensuring the stability of the supply chain. Historical data statistical analysis can be used, combined with the professional judgment of business experts, to determine the relative importance of various types of data. For example, after reviewing multiple beauty product live broadcasts, it is found that transaction flow data directly determines the profitability of the live broadcast, and its proportion in the business may be set at 40%; multimedia stream data is used to display products and attract user attention, with a proportion of 30%; user operation data reflects user interest and participation, with a proportion of 20%; and supply chain status data ensures product supply, with a proportion of 10%. This step solves the problem of difficult to measure the importance of different types of data during data processing and lays a foundation for subsequent targeted data processing and analysis. Next, obtain the actual data volumes corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data. For different types of data, statistics are separately carried out from their storage systems or sources. For multimedia stream data, statistics are made on the total duration of live videos, the number of audio frames, etc.; for user operation data, the number of operation records is counted; for transaction flow data, the number of orders is counted; for supply chain status data, the number of inventory records and transportation track records are counted, etc. For example, in an electronic product live broadcast, the total video duration of the multimedia stream data is counted as 400 minutes, the number of records of user operation data is 1200, the number of orders of transaction flow data is 250, there are 60 inventory records and 40 transportation track records in the supply chain status data. By obtaining the actual data volumes, the scale of each type of data can be intuitively understood, solving the problem of lack of accurate understanding of the data scale, which helps to judge the reliability and representativeness of the data. Then, obtain the data quality evaluation values corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data according to the data acquisition quality evaluation value, data proportion, and actual data volume; the data acquisition quality evaluation value is a comprehensive measure of the quality of the data acquisition process. On this basis, it is adjusted in combination with the data proportion and actual data volume. Data with a high data proportion and a large actual data volume will have a greater influence in the final data quality evaluation. For example, assuming the data acquisition quality evaluation value is 70 points, the proportion of transaction flow data is 40%, the actual number of orders is 250, the proportion of user operation data is 20%, and the actual number of operation records is 1200. By comprehensively considering these factors (weighted calculation can be used, that is, adjusting the acquisition quality evaluation value according to the proportion of the data proportion and actual data volume in the total data volume), the data quality evaluation values of the transaction flow data and user operation data are obtained.This step avoids the one-sidedness of evaluating data quality by a single factor, making the evaluation results more in line with the actual business needs. After that, the total amount of corrected data is obtained based on multimedia stream data, user operation data, transaction flow data, and supply chain status data. Considering various factors such as business experience, historical data, and industry standards, a reasonable total amount of corrected data is set for each type of data. For example, referring to the experience of past live broadcasts of the same type, it is expected that for a 3-hour live broadcast of household items, the video duration of the multimedia stream data should be between 300 and 360 minutes; according to the industry average level and the scale of the live broadcast, the number of recorded user operation data is expected to be between 1000 and 1500; according to market demand and product popularity, the number of orders for transaction flow data is expected to be between 200 and 300; according to the scale of the supply chain and product characteristics, the inventory and transportation records of supply chain status data are expected to be between 60 and 80. These preset values are not fixed and can be dynamically adjusted according to the actual business situation and changes. This step provides a reference standard for evaluating data quality and solves the problem of difficulty in judging the integrity of data collection due to the lack of a reference standard for the amount of data. Finally, a data collection judgment value is obtained based on the total amount of corrected data and the data quality evaluation value. The total amount of corrected data for each type of data is analyzed in combination with the data quality evaluation value. If the data quality evaluation value is high, but the actual amount of data is far lower than the total amount of corrected data, it indicates that although the data quality is good, the quantity is insufficient and further collection may be required; conversely, if the data quality evaluation value is low, even if the actual amount of data reaches the total amount of corrected data, the data may need to be cleaned or re-collected. By comprehensive calculation (for example, combining the proportional relationship between the data quality evaluation value, the actual amount of data, and the total amount of corrected data), a data collection judgment value is obtained and a threshold is set. According to the comparison result of the judgment value and the threshold, the subsequent processing method of the data is determined. For example, if the judgment value is higher than the threshold, it means that the data basically meets the requirements and can enter the next step of analysis or storage; if it is lower than the threshold, corresponding measures need to be taken. This step comprehensively evaluates the effect of data collection and solves the problem of being unable to comprehensively evaluate the data collection effect and making reasonable decisions.

[0035] As Figure 2 shown, the present invention also provides a business product live stream data collection system, including: A first acquisition module 1 for acquiring multiple groups of live source data in a business live broadcast scenario, where each group of the live source data includes multimedia stream data, user operation data, transaction flow data, and supply chain status data; A second acquisition module 2 for obtaining differential data collection cycle information according to the live source data; A third acquisition module 3 for obtaining multiple batches of data sample sets according to the differential data collection cycle information; The fourth acquisition module 4 is used to perform integrated analysis on the multi-batch data sample set to obtain a data acquisition quality evaluation value; The fifth acquisition module 5 is used to obtain a data acquisition judgment value according to the data acquisition quality evaluation value; The judgment module 6 is used to judge whether the data acquisition judgment value is greater than a preset threshold; If it is greater, the live source data is stored in the core database; If it is less, the live source data is excluded.

[0036] In one embodiment, the first acquisition module 1 includes: The first acquisition unit is used to acquire multiple types of data access points in the business live broadcast scenario; The second acquisition unit is used to acquire multiple groups of original data according to the multiple types of data access points; The first judgment unit is used to acquire the format feature of each group of original data and judge the corresponding original data as multimedia stream data according to the format feature; The second judgment unit is used to acquire the user operation feature of each group of original data and judge the corresponding original data as user operation data according to the user operation feature; The third judgment unit is used to acquire the commodity trading feature of each group of original data and judge the corresponding original data as transaction flow data according to the commodity trading feature; The fourth judgment unit is used to acquire the commodity logistics feature of each group of original data and judge the corresponding original data as supply chain status data according to the commodity logistics feature.

[0037] In one embodiment, the second acquisition module 2 includes: The third acquisition unit is used to acquire the data structure in the audio and video database according to the multimedia stream data; The fourth acquisition unit is used to acquire the timestamp feature information of the key frames according to the data structure; The establishment unit is used to establish a key frame index and a scene label according to the timestamp feature information; The fifth acquisition unit is used to acquire the update frequency of the key frame index and the scene label, and use the update frequency as the first acquisition period; The sixth acquisition unit is used to acquire multiple activity levels according to the user operation data and acquire the second acquisition period of the user operation data according to the multiple activity levels; The seventh acquisition unit is used to acquire the business dependence degree of the business decision on the transaction data according to the transaction flow data and the third acquisition period according to the business dependence degree; An eighth acquisition unit, configured to obtain a business-supply chain sensitivity according to the supply chain status data, and obtain a fourth acquisition period according to the business-supply chain sensitivity; A first generation unit, configured to summarize and integrate the first acquisition period, the second acquisition period, the third acquisition period, and the fourth acquisition period to generate differential data acquisition period information.

[0038] In one embodiment, the third acquisition module 3 includes: A ninth acquisition unit, configured to obtain a plurality of data acquisition tasks according to the differential data acquisition period information; A first acquisition unit, configured to obtain acquisition configuration parameters according to each data acquisition task; A tenth acquisition unit, configured to obtain original traceability data sources corresponding to multimedia stream data, user operation data, transaction flow data, and supply chain status data according to the acquisition configuration parameters; A traceability unit, configured to obtain an acquisition sequence according to each of the original traceability data sources, where the acquisition sequence includes an acquisition order, an acquisition frequency, and an acquisition interval; A second acquisition unit, configured to perform high-frequency preliminary acquisition on each original traceability data source according to a preset acquisition sequence to obtain acquisition information; A third acquisition unit, configured to integrate the acquisition information corresponding to the multimedia stream data, the user operation data, the transaction flow data, and the supply chain status data to form a multi-batch data sample set.

[0039] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, value library, or other medium provided in this application and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0040] It should be noted that in this article, the terms "include", "comprise", or any other variant thereof are intended to cover non-exclusively, so that a process, device, article, or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, device, article, or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, device, article, or method including such element.

[0041] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent results or equivalent process transformations made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, are equally included in the patent protection scope of the present invention.

Claims

1. A method for collecting live stream data of business products, characterized in that, Including: Obtain multiple groups of live source data in the business live broadcast scenario, where each group of the live source data includes multimedia stream data, user operation data, transaction flow data, and supply chain status data; Obtain differential data collection cycle information based on the live source data; Obtain multiple batches of data sample sets according to the differential data collection cycle information; Integrate and analyze the multiple batches of data sample sets to obtain a data collection quality evaluation value; Obtain a data collection judgment value according to the data collection quality evaluation value; Judge whether the data collection judgment value is greater than a preset threshold; If it is greater, store the live source data in the core database; If it is less, eliminate the live source data.

2. The method for collecting live stream data of business products according to claim 1, wherein The step of obtaining multiple groups of live source data in the business live broadcast scenario includes: Obtain multiple types of data access points in the business live broadcast scenario; Obtain multiple groups of original data according to the multiple types of data access points; Obtain the format characteristics of each group of original data, and judge the corresponding original data as multimedia stream data according to the format characteristics; Obtain the user operation characteristics of each group of original data, and judge the corresponding original data as user operation data according to the user operation characteristics; Obtain the commodity trading characteristics of each group of original data, and judge the corresponding original data as transaction flow data according to the commodity trading characteristics; Obtain the commodity logistics characteristics of each group of original data, and judge the corresponding original data as supply chain status data according to the commodity logistics characteristics.

3. A method for collecting live stream data of business products according to claim 1, characterized in that, The step of obtaining differential data collection cycle information based on the live source data includes: Obtain the data structure in the audio and video database according to the multimedia stream data; Obtain the timestamp characteristic information of the key frames according to the data structure; Establish a key frame index and a scene label according to the timestamp characteristic information; Obtain the update frequency of the key frame index and the scene label, and use the update frequency as the first collection cycle; Obtain multiple activity levels according to the user operation data, and obtain the second collection cycle of the user operation data according to the multiple activity levels; Obtain the business dependence degree of the business decision on the transaction data according to the transaction flow data, and obtain the third collection cycle according to the business dependence degree; Obtain the business-supply chain sensitivity according to the supply chain status data, and obtain the fourth collection cycle according to the business-supply chain sensitivity; Summarize and integrate the first collection cycle, the second collection cycle, the third collection cycle, and the fourth collection cycle to generate differential data collection cycle information.

4. A method for collecting live stream data of business products according to claim 1, characterized in that, The step of obtaining multiple batches of data sample sets according to the differential data collection cycle information includes: Obtain multiple data collection tasks according to the differential data collection cycle information; Obtain collection configuration parameters according to each data collection task; Obtain the original traceable data sources corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data according to the collection configuration parameters; Obtain a collection sequence according to each of the original traceable data sources, where the collection sequence includes a collection order, a collection frequency, and a collection interval; Perform high-frequency preliminary collection on each original traceable data source according to a preset collection sequence to obtain collection information; Integrate the collected information corresponding to multimedia stream data, user operation data, transaction flow data and supply chain status data to form multiple batches of data sample sets.

5. A method for collecting live stream data of business products according to claim 1, characterized in that, The step of integrating and analyzing the multiple batches of data sample sets to obtain a data collection quality assessment value includes: Acquire multiple batches of data structure formats and multiple batches of data contents according to the multiple batches of data sample sets; Acquire format result information according to the multiple batches of data structure formats, wherein the format result information includes structured data result information, semi-structured data result information and unstructured data result information; Obtaining an information continuity evaluation value according to the structured data result information, the semi-structured data result information, and the unstructured data result information; Acquire content result information according to the content of the multiple batches of data, wherein the content result information includes information completeness, information accuracy, information timeliness, information clarity, information relevance and information continuity; Obtaining content information evaluation values ​​based on the analytic hierarchy process; A data collection quality evaluation value is obtained according to the information continuity evaluation value and the content information evaluation value.

6. The method for collecting live stream data of a business product according to claim 1, wherein The step of obtaining the data collection judgment value according to the data collection quality evaluation value comprises: Acquire the data proportion in the live broadcast business of commercial products according to the multimedia stream data, user operation data, transaction flow data and supply chain status data; Obtaining actual data volumes corresponding to the multimedia stream data, user operation data, transaction flow data, and supply chain status data; Obtain data quality assessment values ​​corresponding to multimedia stream data, user operation data, transaction flow data and supply chain status data according to the data collection quality assessment value, data proportion and actual data volume; Acquire a total correction data volume according to the multimedia stream data, user operation data, transaction flow data and supply chain status data; A data collection judgment value is obtained according to the total correction data amount and the data quality evaluation value.

7. A business product live stream data acquisition system, characterized in that, include: A first acquisition module is used to acquire multiple groups of live source data in a business live broadcast scenario, wherein each group of live source data includes multimedia stream data, user operation data, transaction flow data and supply chain status data; A second acquisition module is used to acquire differentiated data collection period information according to the live source data; A third acquisition module is used to acquire multiple batches of data sample sets according to the differentiated data collection cycle information; A fourth acquisition module is used to integrate and analyze the multiple batches of data sample sets to obtain a data collection quality assessment value; A fifth acquisition module, used to acquire a data acquisition judgment value according to the data acquisition quality evaluation value; A judgment module, used to judge whether the data collection judgment value is greater than a preset threshold; If it is greater, the live source data will be stored in the core database; If it is less than that, the live source data will be removed.

8. The business product live stream data acquisition system according to claim 6, characterized in that, The first acquisition module includes: A first acquisition unit is used to acquire multiple types of data access points in a business live broadcast scenario; A second acquisition unit, used for acquiring multiple groups of original data according to multiple types of data access points; A first judgment unit, configured to obtain a format feature of each group of original data, and judge the corresponding original data as multimedia stream data according to the format feature; The second judgment unit is used to obtain the user operation characteristics of each group of original data, and judge the corresponding original data as user operation data according to the user operation characteristics; The third judgment unit is used to obtain the commodity transaction characteristics of each group of original data, and judge the corresponding original data as transaction flow data according to the commodity transaction characteristics; The fourth judgment unit is used to obtain the commodity logistics characteristics of each group of original data, and judge the corresponding original data as supply chain status data according to the commodity logistics characteristics.

9. The business product live stream data acquisition system according to claim 6, wherein The second acquisition module includes: The third acquisition unit is used to obtain the data structure in the audio and video database according to the multimedia stream data; The fourth acquisition unit is used to obtain the timestamp feature information of the key frames according to the data structure; The establishment unit is used to establish a key frame index and a scene label according to the timestamp feature information; The fifth acquisition unit is used to obtain the update frequency of the key frame index and the scene label, and use the update frequency as the first acquisition period; The sixth acquisition unit is used to obtain multiple activity levels according to the user operation data, and obtain the second acquisition period of the user operation data according to the multiple activity levels; The seventh acquisition unit is used to obtain the business dependence degree of the business decision on the transaction data according to the transaction flow data, and the third acquisition period according to the business dependence degree; The eighth acquisition unit is used to obtain the business-supply chain sensitivity according to the supply chain status data, and obtain the fourth acquisition period according to the business-supply chain sensitivity; The first generation unit is used to summarize and integrate the first acquisition period, the second acquisition period, the third acquisition period and the fourth acquisition period to generate differential data acquisition period information.

10. A business product live stream data collection system according to claim 6, characterized in that, The third acquisition module includes: The ninth acquisition unit is used to obtain multiple data acquisition tasks according to the differential data acquisition period information; The first acquisition unit is used to obtain acquisition configuration parameters according to each data acquisition task; The tenth acquisition unit is used to obtain the original trace data sources corresponding to the multimedia stream data, user operation data, transaction flow data and supply chain status data according to the acquisition configuration parameters; The trace unit is used to obtain an acquisition sequence according to each of the original trace data sources, where the acquisition sequence includes an acquisition order, an acquisition number, and an acquisition interval; The second acquisition unit is used to perform high-frequency preliminary acquisition on each original trace data source according to a preset acquisition sequence to obtain acquisition information; The third acquisition unit is used to integrate the acquisition information corresponding to the multimedia stream data, user operation data, transaction flow data and supply chain status data to form a multi-batch data sample set.

Images