Management service platform system based on big data

Abstract

The invention provides a management service platform system based on big data, and belongs to the technical field of computer systems and information processing, and the system comprises a structure layering module which collects data from different data sources in real time, and carries out the structural layering of the data collected in real time, and obtains a structural layering result; the theme classification module is used for acquiring data contents of the data collected in real time, and performing theme classification based on the data contents to obtain a theme classification result; the label generation module is used for generating group labels based on the structured layering result and the subject classification result; the pushing optimization module is used for analyzing the group label to obtain a group label analysis result, performing accurate pushing based on the group label analysis result, and obtaining real-time feedback of the user for optimization; and the system optimization module is used for performing closed-loop management and optimization on the working processes of all the modules and the real-time feedback of the user on the working processes of all the modules based on a preset algorithm. And the pushing precision and the overall optimization capability of the system are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer systems and information processing technology, and in particular to a management service middle platform system based on big data. Background Art

[0002] With the rapid development of big data technology, all kinds of enterprises and organizations need to process massive amounts of data from multiple data sources and apply these data to accurate service push. However, existing management systems usually lack the ability to effectively classify and process data in real time, especially when facing heterogeneous data sources, it is difficult to accurately classify and generate effective push tags.

[0003] Solutions in the prior art generally use static rules or limited algorithms to perform data classification and push optimization. Data classification is not accurate enough and cannot handle complex, multi-source heterogeneous data. The push process lacks a real-time optimization mechanism for user feedback and cannot form a closed-loop management of data processing and push.

[0004] Therefore, the present invention provides a management service middle platform system based on big data. Summary of the invention

[0005] The present invention provides a management service middle platform system based on big data, which is used to generate group tags through structured stratification and subject classification, and realize accurate push based on group tags. By acquiring and analyzing real-time feedback, the push content is optimized, forming a closed-loop management of data collection, analysis, push and feedback, which improves the accuracy of push and the overall optimization capability of the system.

[0006] The present invention provides a management service middle platform system based on big data, comprising: Structural layering module: collects data from different data sources in real time, and structures and layers the real-time collected data to obtain structured layering results; Theme classification module: obtains the data content of the data collected in real time, and classifies the data collected in real time based on the data content to obtain the theme classification results; Label generation module: Generate group labels based on structured hierarchical results and topic classification results; Push optimization module: Analyze group labels to obtain group label analysis results, accurately push based on group label analysis results, and obtain real-time feedback from users on the pushed content to optimize the push process and obtain push optimization results; System optimization module: Based on the preset algorithm, the working process of the structure stratification module, topic classification module, label generation module, and push optimization module, as well as the user's real-time feedback on the working process of the structure stratification module, topic classification module, label generation module, and push optimization module are closed-loop managed and optimized.

[0007] The present invention provides a management service middle platform system based on big data, with a structured hierarchical module, including: Data analysis unit: collects data from different data sources in real time, analyzes the real-time collected data, obtains data analysis results, and determines the data importance coefficient of each piece of data in the real-time collected data based on the data analysis results: ; ; ; in, is the data importance coefficient of the i-th data, is the originality weight of the i-th data, is the number of conversions of the i-th data in the real-time collected data, is the preset maximum number of conversions for the i-th data. is the historical usage frequency of the ith data in the real-time collected data, is the number of times the i-th data item is used in the t-th period, The preset total time period selected for calculating the historical usage frequency, is the duration of the tth period, is the preset event attenuation factor, and , is the business relevance weight of the ith piece of data in the real-time collected data, is the business relevance coefficient of the ith data in the real-time collected data, is the historical dependence coefficient on the i-th data in business decision making, is the preset weight factor for history dependency, is the dependency coefficient of business scenario changes on the i-th data, is the redundancy coefficient of the i-th data, is the historical variability weight of the i-th data, is the historical change coefficient of the ith data; Data stratification unit: Based on the data importance coefficients of all the data in the real-time collected data, the real-time collected data is structured and stratified to obtain result-based stratification results.

[0008] The present invention provides a management service middle platform system based on big data, a data layering unit, including: Boundary determination subunit: determines the boundary values ​​of each layer of the hierarchy based on a preset data allocation rule; Structural layering subunit: Based on the data importance coefficient of all data and the boundary value of each layer, the data will be structured and layered. The data will be divided into three layers, including: source layer, public layer and application layer.

[0009] The present invention provides a management service middle platform system based on big data, a subject classification module, including: Matching construction unit: constructs a matching expression for each preset data information category based on the preset data information category and the preset expression algorithm; Information extraction unit: matches the information of the data collected in real time based on the matching expression of each preset data information category and the preset matching engine; First classification unit: classify and label the data according to the information of the real-time collected data obtained by matching and the preset classification rules. Each matching type represents a different first theme; The content extraction unit extracts key content of the data in each first topic based on a preset language processing technology; The second classification unit: matches the key content of the data with a preset subject keyword library, and then classifies the data in each first subject into several second subjects in the preset keyword library.

[0010] The present invention provides a management service middle platform system based on big data, a label generation module, including: Preliminary generation unit: Generates preliminary labels based on structured hierarchical results and topic classification results; A first label generation unit: divides the generated preliminary labels into multiple levels according to preset label dimensions, thereby obtaining first group labels, wherein the divided first group labels include: demographic labels, behavioral labels, interest labels, and technical labels; The second label generating unit is configured to obtain user data, perform group analysis on the user data, obtain group analysis results, and generate a second group label based on the group analysis results; Label analysis unit: analyzes the first group label and the second group label to obtain the correlation between different labels; Weight acquisition unit: acquires the label weight of each first group label and second group label based on a preset label information library; Label generation unit: Generates several group labels based on the correlation between all labels and label weights.

[0011] The present invention provides a management service middle platform system based on big data, and a preliminary generation unit, including: Label generation subunit: Generates basic labels based on the structured hierarchical results, and generates specific behavior and event labels based on the topic classification results; Label integration subunit: Integrate basic labels with specific behavior and event labels to obtain preliminary labels.

[0012] The present invention provides a management service middle platform system based on big data, a push optimization module, including: Label analysis unit: analyzes group labels to obtain label types of group labels; Feature acquisition unit: acquires user data and analyzes the user data to obtain user group characteristics; Content extraction unit: matching corresponding group tags based on user group characteristics, and extracting corresponding push content from a preset tag content library based on the tag type corresponding to the matched group tag; Strategy acquisition unit: acquires corresponding push strategies from a preset push strategy library based on user group characteristics; Content push unit: Generates personalized push content based on push strategy and push content for content push; Push adjustment unit: collects user feedback data and dynamically adjusts push content based on the feedback data.

[0013] The present invention provides a management service middle platform system based on big data, and the tag types of group tags include: behavior tags, interest tags, time tags, device tags and location tags.

[0014] Compared with the prior art, the present invention has the following beneficial effects: Generate group tags through structured stratification and subject classification, and achieve accurate push based on group tags. Optimize push content through acquisition and analysis of real-time feedback, forming a closed-loop management of data collection, analysis, push and feedback, improving push accuracy and the overall optimization capability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 It is a structural diagram of a management service middle platform system based on big data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Example 1 The embodiment of the present invention provides a management service middle platform system based on big data, such as Figure 1 As shown, including: Structural layering module: collects data from different data sources in real time, and structures and layers the real-time collected data to obtain structured layering results; Theme classification module: obtains the data content of the data collected in real time, and classifies the data collected in real time based on the data content to obtain the theme classification results; Label generation module: Generate group labels based on structured hierarchical results and topic classification results; Push optimization module: Analyze group labels to obtain group label analysis results, accurately push based on group label analysis results, and obtain real-time feedback from users on the pushed content to optimize the push process and obtain push optimization results; System optimization module: Based on the preset algorithm, the working process of the structure stratification module, topic classification module, label generation module, and push optimization module, as well as the user's real-time feedback on the working process of the structure stratification module, topic classification module, label generation module, and push optimization module are closed-loop managed and optimized.

[0019] In this embodiment, structured stratification is the process of formatting and hierarchically organizing multi-source data to make the data more standardized and structured for subsequent processing. The stratification process may depend on the hierarchical division of data sources, data types or certain attributes. Assuming that the data sources of a retail enterprise include e-commerce platforms, physical store sales records and customer questionnaire feedback, the structured stratification module will hierarchically and structure the data from these different sources by type (such as "user data", "product data" and "sales data"). For example, the information of each customer is structured into data hierarchies such as "name-gender-purchase history-preferences"; In this embodiment, the topic classification is based on the data content, and the data is divided into different topic categories according to its semantic content. For example, the data can be classified according to text, tags or keywords to help analyze and understand the main content of the data. Assume that in the comment data collected on a social media platform, the topic classification module divides the comments into topic categories such as "product feedback", "after-sales service", and "activity promotion"; In this embodiment, the topic classification result is the classification information obtained after the data is subject-classified, indicating which topic categories the data belongs to. This result provides a basis for subsequent tag generation and precise push. For example, in the above-mentioned social media comment example, suppose the topic classification result shows that 90% of the comments are concentrated on "after-sales service" and 10% are concentrated on "product feedback".

[0020] In this embodiment, the group tag is a tag generated based on the structured hierarchical results and the topic classification results, which can summarize the common characteristics or preferences of a specific group. Through the group tag, the system can accurately push content to a specific group. For example, in an e-commerce platform, the group tag can be "young women-lovers of skin care products-frequent followers of activities". This tag is generated based on the user's purchase history, age, gender and preferences, so that the system can recommend new skin care products and discount activities to such users.

[0021] In this embodiment, the group label analysis result is the result obtained after analyzing the group label, revealing the behavior pattern or demand trend of the label group. For example, by analyzing the group label of "young women - love skin care products - frequent followers of activities", the group label analysis result may show that this group is most sensitive to promotional activities, especially activities during holidays. Therefore, the system will focus on pushing promotional information of skin care products during these time periods.

[0022] In this embodiment, the preset algorithm is a calculation method pre-set in the system for data analysis, push optimization or feedback management. For example, the preset algorithm may be a recommendation algorithm based on collaborative filtering, which determines the most suitable push content for the user based on the user's interests and preferences of similar users, such as recommending new products related to products that the user has purchased. In this embodiment, closed-loop management refers to the management process implemented by the system through a series of operations such as data collection, analysis, push, feedback, and optimization, so that the entire process can be continuously improved and optimized. Closed-loop management ensures that user feedback will be incorporated into the push optimization and system optimization process again. For example, when the user's feedback on a certain type of push is low (such as low open rate or low click rate), the system will incorporate this feedback into the closed-loop management process, adjust the algorithm and push content to ensure that the next push is more in line with user preferences. For example, the system may reduce the content that users don't like and push content that they are more interested in.

[0023] The beneficial effects of the above technical solution are: generating group labels through structured stratification and subject classification, and achieving accurate push based on group labels. By acquiring and analyzing real-time feedback, the push content is optimized, forming a closed-loop management of data collection, analysis, push and feedback, which improves the accuracy of push and the overall optimization capability of the system.

[0024] Example 2 The embodiment of the present invention provides a management service middle platform system based on big data, with a structured hierarchical module, including: Data analysis unit: collects data from different data sources in real time, analyzes the real-time collected data, obtains data analysis results, and determines the data importance coefficient of each piece of data in the real-time collected data based on the data analysis results: ; ; ; in, is the data importance coefficient of the i-th data, is the originality weight of the i-th data, is the number of conversions of the i-th data in the real-time collected data, is the preset maximum number of conversions for the i-th data. is the historical usage frequency of the ith data in the real-time collected data, is the number of times the i-th data item is used in the t-th period, The preset total time period selected for calculating the historical usage frequency, is the duration of the tth period, is the preset event attenuation factor, and , is the business relevance weight of the ith piece of data in the real-time collected data, is the business relevance coefficient of the ith data in the real-time collected data, is the historical dependence coefficient on the i-th data in business decision making, is the preset weight factor for history dependency, is the dependency coefficient of business scenario changes on the i-th data, is the redundancy coefficient of the i-th data, is the historical variability weight of the i-th data, is the historical change coefficient of the ith data; Data stratification unit: Based on the data importance coefficients of all the data in the real-time collected data, the real-time collected data is structured and stratified to obtain result-based stratification results.

[0025] In this embodiment, the conversion count refers to the number of times a piece of data is obtained and actually applied or processed, which usually measures the conversion frequency of data in actual business. The higher the conversion count, the more frequently the data is used or called in the business. Suppose there is a user click record data in an online shopping platform. This data is used to update the user's recommendation list, and is also used for advertising push and analysis of popular products. Assuming that the data is called 5 times, the conversion count of the data is 5; In this embodiment, the preset maximum number of conversions is the upper limit set by the system that a certain piece of data can be called within a specific time period. After exceeding this number, the call of the data will be restricted or reduced. This setting is usually used to prevent excessive reliance on a certain piece of data or excessive consumption of resources. For example, in an advertising delivery system, the preset maximum number of conversions for a piece of user's interest tag data is 10 times, that is, the data can be called at most 10 times within a certain period of time. If the system has called this data 10 times, further push may give priority to other interest tag data to avoid over-reliance on a single tag; In this embodiment, the historical usage frequency refers to the number and frequency of a piece of data being called in history. It is usually used to measure the importance and reliability of the data in long-term business. For example, in a logistics management system, the historical usage frequency of "order delivery status" data is high because this data is used every time an order is processed and the status is updated, while the historical usage frequency of "coupon usage record" may be low during non-promotional periods, so its data importance may be relatively low; In this embodiment, the preset event decay factor is used to control the degree of decay of the importance of data over time. As time goes by, the importance of data may gradually weaken, so the decay factor is used to reduce the weight of old data to ensure that the system pays more attention to recent data. For example, in a news recommendation system, the preset event decay factor can ensure that news from a month ago is no longer recommended. The decay factor will reduce the importance of these old news, so that the system will give priority to recommending the latest news content; In this embodiment, the historical dependency coefficient refers to the degree of reliance of the system on a piece of data in decision making. Data with a high historical dependency coefficient indicates that the data was frequently used in past business decisions, which means that it has high stability and reference value in the business. For example, in a financial risk control system, the historical dependency coefficient of the "customer's credit score" data is high because this data is extremely important and frequently used when making decisions. In contrast, the historical dependency coefficient of the "customer's recent purchase preference" data may be low because this type of data does not always have a significant impact on the decision; In this embodiment, the dependency coefficient of business scenario changes on data measures the applicability and degree of dependence of a piece of data on the business under different business scenario changes. A high dependency coefficient indicates that even if the business scenario changes, the data is still critical. For example, in a customer relationship management system (CRM), the dependency coefficient of "customer loyalty" data is high, because whether it is a promotion, customer return visit, or after-sales service, the data is applicable in all scenarios, with wide applicability and high degree of dependence. The dependency coefficient of "seasonal product purchase record" data is low because it may only be used in specific promotion or seasonal activity scenarios.

[0026] The beneficial effect of the above technical solution is: through the structure layering module, the data from multiple data sources are analyzed in real time to determine the importance coefficient of each data. Based on this coefficient, the system can effectively structure the data and improve the organization and availability of the data. This method not only optimizes the management and utilization efficiency of data, but also enhances the support ability of data in business decision-making, making the decision-making process more accurate and faster.

[0027] Example 3 The embodiment of the present invention provides a management service middle platform system based on big data, and a data layering unit, including: Boundary determination subunit: determines the boundary values ​​of each layer of the hierarchy based on a preset data allocation rule; Structural layering subunit: Based on the data importance coefficient of all data and the boundary value of each layer, the data will be structured and layered. The data will be divided into three layers, including: source layer, public layer and application layer.

[0028] In this embodiment, the preset data allocation rule is a criterion or algorithm for how to divide data at different levels, and provides a basis for data stratification based on attributes such as data importance, usage frequency, and business dependency, including multiple standards or thresholds to determine which level each piece of data should belong to. For example, in an e-commerce management system, the preset data allocation rule may include the following criteria: Importance: Data with an importance coefficient exceeding a certain value will be placed in the source layer first, Usage frequency: Highly used data tends to be placed in the public layer for quick access by multiple business units, Real-time requirements: Data that requires real-time updates may be placed directly in the application layer to facilitate direct calls by the business system; In this embodiment, the boundary value is the numerical boundary used to define the data hierarchy in the allocation rule. The boundary value of each level is set based on the importance coefficient of the data, the frequency of use and other parameters to ensure that the system can automatically divide the data into appropriate levels according to these boundaries. Assuming that the system uses the importance coefficient of the data (in the range of 0-1) as the main criterion for stratification, the boundary value of each level can be set as follows: Source layer boundary value: data with an importance coefficient above 0.8 is allocated to the source layer, public layer boundary value: data with an importance coefficient between 0.4 and 0.8 is allocated to the public layer, and application layer boundary value: data with an importance coefficient below 0.4 is allocated to the application layer. The setting of boundary values ​​ensures that data enters different levels in the order of high, medium and low importance to meet different data access and processing requirements.

[0029] In this embodiment, the source layer mainly stores the most original and timely data to ensure that the data is quickly acquired and updated by the system without any processing. The design of the source layer is usually to ensure the authenticity and real-time nature of the data and support the system to perform instant analysis or updates. For example, in a logistics system, real-time GPS positioning data will be placed in the source layer because these data are unprocessed original information with high real-time requirements and are directly derived from logistics equipment. In this embodiment, the public layer is a data sharing layer, and the stored data has been preliminarily processed or standardized so that each business module can share access more efficiently. The data in the public layer is suitable for common and frequent use in multiple scenarios and does not need to be updated all the time. For example, in the management system of a retail enterprise, the basic information of users (such as name, gender, and contact information) will be placed in the public layer because these data are basic data used by different business departments (such as the marketing department and the customer service department); In this embodiment, the application layer is the final application layer of the data. The stored data is usually subjected to multiple analyses and processing and has a high business relevance. The data in the application layer is used to support direct calls and decisions of specific applications or business scenarios. For example, in the customer management system of a bank, the "high-value customer" label data generated by the personalized recommendation model will be stored in the application layer, because this type of data is the result of multiple calculations of the model and has a high business specificity, and is mainly used by the marketing department in specific promotion activities.

[0030] The beneficial effects of the above technical solution are: through the data stratification unit, the data is structured and stratified according to the data importance coefficient and boundary value, forming a three-level architecture of the source layer, the public layer and the application layer, optimizing the data storage and access path, improving the data management efficiency and hierarchical utilization capability, and helping to improve the system's response speed and data processing performance in a big data environment.

[0031] Example 4 The embodiment of the present invention provides a management service middle platform system based on big data, a subject classification module, including: Matching construction unit: constructs a matching expression for each preset data information category based on the preset data information category and the preset expression algorithm; Information extraction unit: matches the information of the data collected in real time based on the matching expression of each preset data information category and the preset matching engine; First classification unit: classify and label the data according to the information of the real-time collected data obtained by matching and the preset classification rules. Each matching type represents a different first theme; The content extraction unit extracts key content of the data in each first topic based on a preset language processing technology; The second classification unit: matches the key content of the data with a preset subject keyword library, and then classifies the data in each first subject into several second subjects in the preset keyword library.

[0032] In this embodiment, the preset data information categories refer to some data types or information categories pre-defined by the system to help the system identify and classify data. For example, according to the business needs of the system, the data may be divided into different categories such as "user behavior", "transaction information", "product information", etc. In the e-commerce platform, the preset data information categories may include "user search history", "browsing history", "purchase history" and "evaluation content". These categories help the system quickly identify the specific type of data, which is convenient for subsequent classification and processing; In this embodiment, the preset expression algorithm refers to an algorithm or method for matching data to a certain category, generating a matching expression through specific rules, helping the system to construct an expression form for each data information category, so that the data can be automatically adapted to the corresponding category. For example, in a content review system, the preset expression algorithm may use keyword extraction, pattern matching and other methods to construct matching rules for different behavioral data categories. For example, data "containing transaction keywords (such as purchase, order and price fields)" can generate a matching expression for the "transaction information" category through an algorithm; In this embodiment, the matching expression is a specific rule generated by a preset expression algorithm, which is used to define the characteristics of each data information category. For example, in a social media data analysis system, the matching expression of the "sentiment analysis" category may be "content containing emoticons and emotional vocabulary", such as text with "like", "hate" or emoticons will be classified into the "sentiment analysis" category.

[0033] In this embodiment, the information of the real-time collected data obtained by matching includes: client error (4xx error code): common matches: such as 404, 403, 401 and other error codes, classification topics: after matching the 4xx error code, the system marks it as a client error, example scenario: 404 error (page not found) is usually classified as the "resource unreachable" topic, the system can further analyze the URL where the error occurs and related user behaviors, server error (5xx error code), common matches: such as 500, 502, 503, 504 and other error codes, classification topics: when matching the 5xx error code, the system marks it as a server error, example scenario: 500 Internal Server Error indicates that an unexpected error occurred inside the server, the system can classify these logs as the "server failure" topic, and further analyze the cause of the error.

[0034] In this embodiment, the preset matching engine is a module within the system for identifying and classifying data based on matching expressions. The matching engine is responsible for scanning the data one by one and classifying it into eligible data information categories. For example, in an intelligent customer service system, the preset matching engine can be based on keyword matching, regular expressions and other technologies, by querying and analyzing customer messages to determine whether the message belongs to categories such as "after-sales issues" or "product consultation".

[0035] In this embodiment, the preset classification rules are criteria for further subdividing the data classification, including multiple attributes or conditions, to ensure that the data can be more accurately classified into specific categories or topics. For example, in a bank's risk management system, the classification rules may include "amount exceeds a certain limit", "multiple withdrawals within a specific time period", etc. Data that meets these rules is marked as "high-risk transactions" to facilitate the risk control team to focus on monitoring; In this embodiment, the matching type refers to the type of data classified after being identified according to the matching expression. Different matching types correspond to different data information categories or attribute labels, representing the preliminary subject attribution of the data. For example, in a content recommendation system, if the user's behavior data is identified as "interest data" through a matching expression, its matching type can be "sports", "technology", etc.; In this embodiment, the first topic is the main category to which the data belongs after preliminary classification, indicating the basic topic or category of the data content, which is usually assigned by the first classification unit according to the matching type and classification rules, helping the system to quickly divide the data into topics. For example, in the user behavior analysis of the e-commerce platform, the user's browsing data may be preliminarily classified as "product browsing", and its first topic is "browsing behavior". Other behaviors such as "add to shopping cart" can be marked as the first topic of "shopping behavior"; In this embodiment, the second topic is a further subdivision based on the first topic. By matching key content with the topic keyword library to form a more refined classification, the second topic provides a deeper semantic division to facilitate precise business applications. For example, in a news recommendation system, the first topic may be "news", and the second topic is further divided into "technology news", "sports news", "international news", etc., so that the system can recommend more detailed content based on user preferences.

[0036] The beneficial effects of the above technical solution are: through the multi-level classification mechanism of the subject classification module, accurate subject classification of real-time data is achieved, and the data can be dynamically classified into different topics by using matching expressions, classification rules and keyword libraries, and key content can be extracted through language processing technology to achieve more detailed secondary classification, thereby improving classification accuracy and system data analysis depth, and providing more targeted and hierarchical support for big data analysis.

[0037] Example 5 The embodiment of the present invention provides a management service middle platform system based on big data, a label generation module, including: Preliminary generation unit: Generates preliminary labels based on structured hierarchical results and topic classification results; A first label generation unit: divides the generated preliminary labels into multiple levels according to preset label dimensions, thereby obtaining first group labels, wherein the divided first group labels include: demographic labels, behavioral labels, interest labels, and technical labels; The second label generating unit is configured to obtain user data, perform group analysis on the user data, obtain group analysis results, and generate a second group label based on the group analysis results; Label analysis unit: analyzes the first group label and the second group label to obtain the correlation between different labels; Weight acquisition unit: acquires the label weight of each first group label and second group label based on a preset label information library; Label generation unit: Generates several group labels based on the correlation between all labels and label weights.

[0038] In this embodiment, the preliminary label is a label generated based on the preliminary analysis and classification of the data. It is the first-level identification of data features, reflects the basic characteristics of the data, and helps lay the foundation for subsequent more detailed label generation. For example, in e-commerce user data, if a user's behavior includes browsing the "sports shoes" page multiple times, the system may generate "sports shoes interest" as a preliminary label for the user.

[0039] In this embodiment, the preset label dimensions are different dimensions or categories of classification labels preset by the system, which are used to divide the preliminary labels into multiple levels, including demographics, behavioral characteristics, interest preferences, technical attributes, etc. For example, in an advertising delivery system, the preset label dimensions may include "age", "gender", "purchase behavior", "browsing interests", etc. These dimensions make the label division more hierarchical and facilitate accurate positioning of users; In this embodiment, multi-level division refers to further refining the preliminary label into multiple levels and dimensions to generate labels with more refined classifications, so as to accurately reflect the complexity of the data. For example, in social media, if a user's preliminary label is "travel interest", the system can further divide it based on the "travel" label, for example, into refined labels such as "like natural scenery" and "like city travel", forming multi-level labels; In this embodiment, the first group label is a label category generated based on a preset label dimension. The preliminary label is subdivided by the preset dimension to obtain the first layer of group labels including demographic labels, behavior labels, interest labels and technical labels. For example, in retail data analysis, the first group label may include "age: 20-30 years old", "gender: female", "interest: fitness", "device type: smart phone", etc. These labels can be used to preliminarily divide users into characteristic groups.

[0040] In this embodiment, the second group label is a label generated based on further group analysis of user data, combined with the specific group behavior and characteristics of users, reflecting more targeted label information. For example, in a video recommendation system, if a large number of users watch a certain type of video at night, the system can generate "nighttime viewing preference" for these users as the second group label, which indicates that the behavior pattern of this group has more detailed characteristics.

[0041] In this embodiment, the preset tag information library is a database of all possible tags and their related weights, classification levels, etc. stored by the system. The tag information library provides basic information for tag generation, weight allocation, etc. For example, in an advertising recommendation platform, the tag information library may contain tag information such as "city of residence", "age group", "interest preference", "device type", etc., so that the system can refine user portraits based on these tags and provide personalized recommendations; In this embodiment, the tag weight refers to the importance or priority of each tag in user profiling or data analysis. The size of the weight is preset by the system or dynamically calculated, and directly affects the ranking and recommendation effect of the final generated tags. For example, in an e-commerce recommendation system, if the weight of the "age" tag is 0.7 and the weight of the "interest" tag is 0.3, it means that the age feature is more important than the interest preference. The system will give priority to the user's age feature and recommend products suitable for their age group.

[0042] In this embodiment, the group label is the final generated label set, which is a comprehensive label formed by combining the first group label and the second group label after weight and correlation calculation, and is used to accurately describe the user or data group. For example, in a bank customer management system, the group label generated based on the user's consumption, income, age, region and other data may be "high net worth customers", "young family users", etc. These group labels help banks customize corresponding service strategies.

[0043] The beneficial effects of the above technical solution are: through the multi-level label generation module, in-depth analysis and accurate classification of user data are achieved. The preliminary labels are subdivided into demographic, behavioral, interest and technical labels, and further combined with the group analysis results to form richer group labels. This method not only improves the relevance and weight evaluation of labels, but also enhances the system's responsiveness in complex data environments, making personalized services and market insights more accurate, and providing strong data support for decision-making.

[0044] Example 6 The embodiment of the present invention provides a management service middle platform system based on big data, and a preliminary generation unit, including: Label generation subunit: Generates basic labels based on the structured hierarchical results, and generates specific behavior and event labels based on the topic classification results; Label integration subunit: Integrate basic labels with specific behavior and event labels to obtain preliminary labels.

[0045] In this embodiment, basic tags include: user behavior data: generated tags include "mobile user", "PC user", "high-frequency access user", "low-activity user", etc., device data: tags are "iOS user", "Android user", "desktop user", etc., log data: generate basic tags such as "HTTP 500 error", "CPU high occupancy", "insufficient memory", etc., identify system performance and errors, for example, a user accesses the system through a mobile terminal, his behavior data will be classified as "mobile user", a server log detects that the CPU occupancy rate is high for a long time, and the system will generate a "CPU high occupancy" tag; In this embodiment, the specific behavior and event tags are based on the subject classification, and the system generates behavior tags or event tags that reflect the data content. These tags describe the actual activities or events in the data. Purchase behavior: the system generates tags such as "high-spending users", "first-time purchasing users", "frequent purchasing users", etc. Browsing behavior: generates tags such as "long-time browsing users", "commodity interest users", etc. System logs: generate tags such as "frequent server abnormalities", "high network latency", "security warnings" for log data. For example, if a user frequently purchases electronic products on the mobile terminal, the system can generate "high-spending users" and "electronic product preference users" tags. If a server often has 500 errors, the system will generate a "frequent server abnormalities" tag.

[0046] The beneficial effects of the above technical solution are: through the collaborative work of label generation and integration sub-units, preliminary labels containing basic, behavioral and event information are generated, multi-level and multi-dimensional data labeling processing is realized, the accuracy of label generation and the recognition ability of the system are improved, and a more detailed and comprehensive label foundation is laid for subsequent data analysis and precise push, which enhances the intelligence of the system and the refinement of data processing.

[0047] Example 7 The embodiment of the present invention provides a management service middle platform system based on big data, a push optimization module, including: Label analysis unit: analyzes group labels to obtain label types of group labels; Feature acquisition unit: acquires user data and analyzes the user data to obtain user group characteristics; Content extraction unit: matching corresponding group tags based on user group characteristics, and extracting corresponding push content from a preset tag content library based on the tag type corresponding to the matched group tag; Strategy acquisition unit: acquires corresponding push strategies from a preset push strategy library based on user group characteristics; Content push unit: Generates personalized push content based on push strategy and push content for content push; Push adjustment unit: collects user feedback data and dynamically adjusts push content based on the feedback data.

[0048] In this embodiment, user group characteristics are characteristic information obtained by analyzing user data, which are used to describe the common characteristics of a certain user group, such as age, region, hobbies, consumption habits, etc. For example, on a music streaming platform, user group characteristics may include "age group: 20-30 years old", "preferred music type: pop, rock", "high frequency usage time: evening". These characteristics help the platform recommend appropriate music content, such as pushing popular playlists in the evening.

[0049] In this embodiment, the preset tag content library is a content database that the system stores according to tag classification, and contains push content corresponding to each tag type. The content library provides materials for personalized push, ensuring that the pushed content matches the user's tag. For example, in an e-commerce platform, the tag content library may include "new product recommendations", "holiday promotions", "limited-time discounts" and other content, and is classified according to tag type. During the "Double Eleven" period, the system can push limited-time discount product information to users based on the "promotional activities" tag.

[0050] In this embodiment, the preset push strategy library is a collection of different push strategies stored in the system. The push method, frequency, time, etc. are determined according to the user's group characteristics and usage scenarios to ensure the best push effect. For example, in a news application, the push strategy library may contain strategies such as "real-time hot spot push" and "daily morning headline push". For users with the "Follow current affairs" tag, the system will push today's news at 7 a.m. to ensure that the push is within the user's active time.

[0051] In this embodiment, personalized push content refers to customized push information generated based on user group characteristics and push strategies, so that the content received by each user is more personalized and relevant, improving user experience and interaction rate. For example, on a video recommendation platform, based on the user's viewing preferences and interest tags, the system generates "personalized push content" such as recommending comedy movies, works by specific directors, etc., to ensure that the pushed content is in line with the user's viewing preferences.

[0052] In this embodiment, the user's feedback data is the user's reaction information after receiving the pushed content, such as click-through rate, likes, comments, shares, etc. The feedback data reflects the user's interest and acceptance of the pushed content, and is used to optimize and adjust the subsequent push strategy. For example, in a shopping application, if the user frequently clicks or collects a certain type of product recommendation, the feedback data records these operations, and the system will increase the frequency of recommending this type of product; if a certain type of content has no response for a long time, the push frequency of this content will be reduced.

[0053] The beneficial effects of the above technical solution are: accurate push is achieved through multi-module collaboration, and push content is dynamically adjusted based on tag type, user group characteristics and push strategy. The push effect is optimized in real time using user feedback data, making the content more personalized and accurate, improving user interaction and response rate, meeting user personalized needs while improving the effectiveness and accuracy of push, and combining big data analysis with feedback mechanism to make the push system more flexible and adaptable.

[0054] Example 8 An embodiment of the present invention provides a management service middle platform system based on big data, and the tag types of group tags include: behavior tags, interest tags, time tags, device tags and location tags.

[0055] In this embodiment, the behavior tag is to identify the user's operation behavior, such as "high-frequency access user", "low-activity user", "shopping user"; In this embodiment, the interest tag is to identify the user's interest area, such as "electronic product preference", "travel enthusiast", "sports enthusiast"; In this embodiment, the timeliness tag is a time attribute that identifies the user behavior, such as "Double 11 high-spending user" and "festival active user"; In this embodiment, the device tag is to identify the type of device or channel used by the user, such as "mobile user", "PC user", "iOS user"; In this embodiment, the location tag identifies the user's geographical location or permanent residence, such as "urban user" or "rural user".

[0056] The beneficial effects of the above technical solution are: through multi-dimensional group tag types, comprehensive analysis and accurate identification of user characteristics can be achieved, and by integrating different tags, user needs and behavior patterns can be better understood, providing more personalized and timely services. This innovative method improves data utilization efficiency, enhances user experience, and enables the service middle-end system to quickly adapt to user changes in a dynamic market, improving overall management and service capabilities.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A management service middle platform system based on big data, characterized in that: include: Structural layering module: collects data from different data sources in real time, and structures and layers the real-time collected data to obtain structured layering results; Theme classification module: obtains the data content of the data collected in real time, and classifies the data collected in real time based on the data content to obtain the theme classification results; Label generation module: Generate group labels based on structured hierarchical results and topic classification results; Push optimization module: Analyze group labels to obtain group label analysis results, accurately push based on group label analysis results, and obtain real-time feedback from users on the pushed content to optimize the push process and obtain push optimization results; System optimization module: Based on the preset algorithm, the working process of the structure stratification module, topic classification module, label generation module, and push optimization module, as well as the user's real-time feedback on the working process of the structure stratification module, topic classification module, label generation module, and push optimization module are closed-loop managed and optimized.

2. According to the big data-based management service middle platform system of claim 1, it is characterized in that: Structural layered modules, including: Data analysis unit: collects data from different data sources in real time, analyzes the real-time collected data, obtains data analysis results, and determines the data importance coefficient of each piece of data in the real-time collected data based on the data analysis results: ； ； ； in, is the data importance coefficient of the i-th data, is the originality weight of the i-th data, is the number of conversions of the i-th data in the real-time collected data, is the preset maximum number of conversions for the i-th data. is the historical usage frequency of the ith data in the real-time collected data, is the number of times the i-th data item is used in the t-th period, The preset total time period selected for calculating the historical usage frequency, is the duration of the tth period, is the preset event attenuation factor, and , is the business relevance weight of the ith piece of data in the real-time collected data, is the business relevance coefficient of the ith data in the real-time collected data, is the historical dependence coefficient on the i-th data in business decision making, is the preset weight factor for history dependency, is the dependency coefficient of business scenario changes on the i-th data, is the redundancy coefficient of the i-th data, is the historical variability weight of the i-th data, is the historical change coefficient of the ith data; Data stratification unit: Based on the data importance coefficients of all the data in the real-time collected data, the real-time collected data is structured and stratified to obtain result-based stratification results.

3. According to the big data-based management service middle platform system of claim 2, it is characterized in that: Data hierarchical unit, including: Boundary determination subunit: determines the boundary values ​​of each layer of the hierarchy based on a preset data allocation rule; Structural layering subunit: Based on the data importance coefficient of all data and the boundary value of each layer, the data will be structured and layered. The data will be divided into three layers, including: source layer, public layer and application layer.

4. According to the big data-based management service middle platform system of claim 1, it is characterized in that: Theme classification module, including: Matching construction unit: constructs a matching expression for each preset data information category based on the preset data information category and the preset expression algorithm; Information extraction unit: matches the information of the data collected in real time based on the matching expression of each preset data information category and the preset matching engine; First classification unit: classify and label the data according to the information of the real-time collected data obtained by matching and the preset classification rules. Each matching type represents a different first theme; The content extraction unit extracts key content of the data in each first topic based on a preset language processing technology; The second classification unit: matches the key content of the data with a preset subject keyword library, and then classifies the data in each first subject into several second subjects in the preset keyword library.

5. According to the big data-based management service middle platform system of claim 1, it is characterized in that: Label generation module, including: Preliminary generation unit: Generates preliminary labels based on structured hierarchical results and topic classification results; A first label generation unit: divides the generated preliminary labels into multiple levels according to preset label dimensions, thereby obtaining first group labels, wherein the divided first group labels include: demographic labels, behavioral labels, interest labels, and technical labels; The second label generating unit is configured to obtain user data, perform group analysis on the user data, obtain group analysis results, and generate a second group label based on the group analysis results; Label analysis unit: analyzes the first group label and the second group label to obtain the correlation between different labels; Weight acquisition unit: acquires the label weight of each first group label and second group label based on a preset label information library; Label generation unit: Generates several group labels based on the correlation between all labels and label weights.

6. A management service middle platform system based on big data according to claim 5, characterized in that: Initial generation unit, including: Label generation subunit: Generates basic labels based on the structured hierarchical results, and generates specific behavior and event labels based on the topic classification results; Label integration subunit: Integrate basic labels with specific behavior and event labels to obtain preliminary labels.

7. According to the big data-based management service middle platform system of claim 1, it is characterized in that: Push optimization module, including: Label analysis unit: analyzes group labels to obtain label types of group labels; Feature acquisition unit: acquires user data and analyzes the user data to obtain user group characteristics; Content extraction unit: matching corresponding group tags based on user group characteristics, and extracting corresponding push content from a preset tag content library based on the tag type corresponding to the matched group tag; Strategy acquisition unit: acquires corresponding push strategies from a preset push strategy library based on user group characteristics; Content push unit: Generates personalized push content based on push strategy and push content for content push; Push adjustment unit: collects user feedback data and dynamically adjusts push content based on the feedback data.

8. According to the big data-based management service middle platform system of claim 1, it is characterized in that: The tag types of group tags include: behavior tags, interest tags, time tags, device tags, and location tags.

Images