User value mining method and system based on dynamic protocol analysis and multi-modal AI

Abstract

The invention relates to the technical field of communication big data analysis and user behavior recognition, in particular to a user value mining method and system based on dynamic protocol analysis and multi-modal AI, and the method comprises the following steps: data collection and primary processing; analyzing a dynamic protocol; carrying out AI intelligent identification; an OneID account number graph is constructed; user value mining and precise service; the method has the advantages that the improved DPI secondary deep analysis technology is combined with the AI algorithm, multilevel feature extraction is carried out on original code stream data of the fixed mobile network, a VIP rule recognition library and a dynamically updated crawler knowledge library are constructed, and full-dimensional recognition of user behavior features is achieved. A mobile phone number and an internet account number are associated by adopting an OneID system, and an XDR ticket is combined for comparative analysis, so that the recognition accuracy rate reaches 96% or above, and 10000 + mainstream APP protocols are covered. According to the method, the problems that the traditional rule identification coverage rate is low, and the social relationship such as the family circle / friend circle cannot be associated are effectively solved, and data support is provided for precision marketing.

Description

Technical Field

[0001] This invention relates to the field of communication big data analysis and user behavior recognition technology, specifically to a user value mining method and system based on dynamic protocol parsing and multimodal AI. Background Technology

[0002] Current user value mining technologies used by telecom operators primarily rely on traditional DPI parsing and static rule bases. These technologies fall into three categories: 1. Basic Protocol Parsing: This involves matching URLs, cookies, and other features using a fixed rule base to identify approximately 3,000 basic protocols (such as HTTP / 1.1). Offline batch processing is used to analyze XDR call detail records (CDRs) and tag user behavior (e.g., "video user"). 2. User Profile Modeling: This relies on manually defined rules to determine high-value users, such as monthly data usage thresholds or simply associating WiFi MAC addresses with internet accounts. Cross-platform association accuracy is less than 60%. 3. Benefits Operation System: This system relies on single membership tags to push benefits, lacks a dynamic update mechanism, and cannot identify the VIP account lifecycle (level, validity period) and family / social relationships.

[0003] Traditional technical methods can lead to the following problems in content parsing and user rights:

[0004] Insufficient protocol coverage: HTTPS encrypted content parsing rate <80%, coverage rate for emerging apps (TikTok / Xiaohongshu) <40% (static rules update slowly, unable to dynamically recognize changes in VIP account levels (e.g., video membership upgrade from VIP1 to VIP6)).

[0005] Low recognition accuracy: VIP level misjudgment rate >30%, family circle misjudgment rate >40% (lacking AI time series analysis and social relationship modeling).

[0006] Data fragmentation: Cross-platform account association accuracy is less than 60% (relying on a single dimension and not building a multi-source integrated OneID system). For example, mobile phone numbers and internet accounts (such as Taobao IDs and WeChat accounts) lack deep association, making it difficult to build a family social relationship graph. Summary of the Invention

[0007] The purpose of this invention is to provide a user value mining method and system based on dynamic protocol parsing and multimodal AI to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a user value mining method based on dynamic protocol parsing and multimodal AI, comprising the following steps:

[0009] Data collection and preliminary processing;

[0010] Dynamic protocol parsing;

[0011] AI intelligent recognition;

[0012] OneID account graph construction;

[0013] User Value Mining and Precision Service: By combining innovative and optimized log analysis models, we can deeply mine user behavior patterns and preference characteristics across different platforms to improve the accuracy of key information identification; we can build a dynamically updated user value assessment system, establish personalized demand models through multi-dimensional data analysis, and achieve precise benefit recommendations based on user profiles to improve service matching and commercial value conversion.

[0014] Preferably, the specific operations for data collection and preliminary processing are as follows: deeply analyze the raw DPI bitstream data to accurately extract the core content accessed by users; combine intelligent algorithms to extract data features and behavioral patterns.

[0015] Preferably, the specific operations of dynamic protocol parsing are as follows: An automated, end-to-end DPI-enhanced technology is constructed to track the protocol interaction process of business applications; based on fixed-mobile converged DPI deep parsing technology, a multi-layered protocol identification system is built to achieve end-to-end tracking of mainstream APP traffic characteristics and load content mining; through automated protocol interaction tracking, deep packet load detection, and XDR call detail record comparison analysis, key fields such as host, url, and cookie are extracted, and combined with connection patterns and traffic statistics rules, multiple protocol types and business scenarios are accurately identified; web crawling capabilities are introduced to capture public network content entities in real time based on URL resource attribute IDs, dynamically update the parsing rule base, achieve dynamic translation capabilities for over 90% of resources, cover popular short videos and celebrity fan information, and improve the real-time performance and coverage of content parsing.

[0016] Preferably, the specific operations of AI intelligent recognition are as follows: relying on AI algorithms to perform deep learning on payload data to achieve multi-dimensional mining and secondary analysis of protocol features; extracting the original payload structure and business features, and combining the crawler verification mechanism to accurately identify user VIP level, resource preferences, social relationships and search browsing behavior, and build a VIP member rule base and resource ID recognition base; introducing a self-learning optimization mechanism, integrating manual intervention and automatic hot patch updates to achieve full lifecycle management of member features, accurately identify high-value rights users, and support refined operations.

[0017] The preferred method for constructing the OneID account graph is as follows: Integrating the content knowledge base and member database to build a unified knowledge system across the entire network, covering resources across all scenarios including video, short video, music, reading, and e-commerce, achieving a 96% HTTP call detail record (CDR) recognition rate, and supporting multi-dimensional parsing of apps, URLs, and keywords; forming a network-wide account graph centered on OneID through strong association between mobile phone numbers and internet accounts, connecting multi-layered relationship chains such as user groups, family circles, and social circles; and combining dynamic integration of the crawler platform and the DPI engine to achieve fusion analysis of hundreds of millions of resource tags, providing underlying capability support for accurate user profiling and cross-business correlation analysis.

[0018] A system for user value mining based on dynamic protocol parsing and multimodal AI includes the following modules:

[0019] Data acquisition and preliminary processing module: used for in-depth analysis of DPI raw bitstream data, accurate extraction of core user access content; and combined with intelligent algorithms to extract data features and behavioral patterns;

[0020] Dynamic Protocol Parsing Module: This module features enhanced DPI technology to automate the entire process of tracking protocol interactions in business applications. Based on fixed-mobile converged DPI deep parsing technology, it constructs a multi-layered protocol identification system, enabling full-process tracking and load content mining of mainstream app traffic characteristics. Through automated protocol interaction tracking, deep packet load detection, and XDR call detail record comparison analysis, it extracts key fields such as host, URL, and cookie. Combined with connection patterns and traffic statistics rules, it accurately identifies multiple protocol types and business scenarios. It also incorporates web crawling capabilities to capture public network content entities in real time based on URL resource attribute IDs, dynamically updating the parsing rule base to achieve dynamic translation capabilities for over 90% of resources, covering popular short videos and celebrity fan information, thus improving the real-time performance and coverage of content parsing.

[0021] AI Intelligent Recognition Module: Relying on AI algorithms to perform deep learning on payload data, it achieves multi-dimensional mining and secondary analysis of protocol features; it extracts the original payload structure and business features, and combines a web crawler verification mechanism to accurately identify user VIP levels, resource preferences, social relationships, and search and browsing behaviors, and builds a VIP member rule base and resource ID recognition base; it introduces a self-learning optimization mechanism, integrates manual intervention and automatic hot patch updates, realizes full lifecycle management of member features, accurately identifies high-value users, and supports refined operations;

[0022] OneID Account Graph Construction Module: Integrates content knowledge base and member database to build a unified knowledge system across the entire network, covering resources across all scenarios including video, short video, music, reading, and e-commerce. It achieves a 96% HTTP call detail record recognition rate and supports multi-dimensional parsing of APP / URL / keywords. Through the strong association between mobile phone numbers and internet accounts, it forms a network-wide account graph centered on OneID, connecting multi-layered relationship chains of user groups, family circles, and social circles. Combined with the dynamic connection between the crawler platform and the DPI engine, it enables the fusion analysis of hundreds of millions of resource tags, providing underlying capability support for accurate user profiling and cross-business correlation analysis.

[0023] User Value Mining and Precision Service Module: Combining innovative and optimized log parsing models, this module deeply mines user behavior patterns and preference characteristics across different platforms, improving the accuracy of key information identification; it constructs a dynamically updated user value assessment system, establishes personalized demand models through multi-dimensional data analysis, and achieves precise benefit recommendations based on user profiles, improving service matching and commercial value conversion.

[0024] Preferably, in the data acquisition and preliminary processing module, fusion verification and multi-dimensional information comparison are performed by crawling to overcome the DPI recognition problem in complex network environments.

[0025] Preferably, in the AI ​​intelligent recognition module, the AI ​​algorithm is a deep neural network algorithm, which trains on a large amount of payload data to achieve multi-dimensional mining and secondary analysis of protocol features, thereby improving the recognition accuracy of user behavior and business features.

[0026] Preferably, in the OneID account graph construction module, the content knowledge base and member database are integrated using data fusion technology to uniformly process and store data from different sources and in different formats, thereby constructing a unified knowledge system across the entire network.

[0027] Preferably, in the user value mining and precision service module, the dynamically updated user value assessment system is dynamically adjusted based on users' historical behavior data, real-time behavior data, and business scenario data to ensure the accuracy and timeliness of the assessment results.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] This invention proposes a user value mining method and system based on dynamic protocol parsing and multimodal AI. Through improved DPI secondary deep parsing technology combined with AI algorithms, it extracts multi-level features from raw fixed-mobile network bitstream data, constructs a VIP rule recognition library and a dynamically updated crawler knowledge base, achieving full-dimensional recognition of user behavior characteristics. Using the OneID system to associate mobile phone numbers with internet accounts, combined with XDR call detail record (CDR) comparative analysis, the recognition accuracy reaches over 96%, covering protocols of over 10,000 mainstream apps. This invention effectively solves the problems of low coverage and inability to associate with social relationships such as family / friends circles in traditional rule recognition, providing data support for precision marketing. Attached Figure Description

[0030] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the present invention clear and complete, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only some, not all, embodiments of the present invention, and are merely illustrative of the embodiments of the present invention. They are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Example 1: This invention provides a technical solution: a user value mining method based on dynamic protocol parsing and multimodal AI, comprising the following steps:

[0033] Based on a big data analytics platform, a full-chain user insight and precise service system is established to extract value from operator users. This is achieved through the following five steps:

[0034] Step 1: Accurately extract the core content accessed by users by deeply analyzing the raw DPI bitstream data;

[0035] Step 2: Extract data features and behavioral patterns using intelligent algorithms;

[0036] Step 3: Overcome the challenge of DPI recognition in complex network environments by using web crawling for fusion verification and multi-dimensional information comparison.

[0037] Step 4: Combine innovative and optimized log parsing models to deeply mine user behavior patterns and preference characteristics across different platforms, thereby improving the accuracy of key information identification.

[0038] Step 5: Build a dynamically updated user value assessment system, establish a personalized demand model through multi-dimensional data analysis, and ultimately achieve accurate benefit recommendations based on user profiles, effectively improving service matching and commercial value conversion.

[0039] The implementation process of specific technical points:

[0040] The specific implementation process of "dynamic protocol parsing":

[0041] Step 1: Build an automated, end-to-end DPI-enhanced technology to track the protocol interaction process of business applications.

[0042] Step 2: Based on fixed-mobile fusion DPI deep analysis technology, construct a multi-layer protocol identification system to achieve full-process tracking and load content mining of traffic characteristics of mainstream apps (video / short video / music / reading / e-commerce, etc.).

[0043] Step 3: Through automated protocol interaction tracking, deep packet load detection, and XDR call detail record comparison analysis, extract key fields such as host, URL, and cookie. Combine connection patterns and traffic statistics rules to accurately identify multiple protocol types and business scenarios.

[0044] Step 4: Introduce web crawling capabilities to capture public network content entities in real time based on URL resource attribute IDs, dynamically update the parsing rule base, achieve dynamic translation capabilities for more than 90% of resources, cover popular short videos, celebrity fan information, etc., and improve the real-time performance and coverage of content parsing.

[0045] The specific implementation process of "AI intelligent recognition":

[0046] Step 1: Relying on AI algorithms to perform deep learning on payload data, we can achieve multi-dimensional mining and secondary analysis of protocol features.

[0047] Step 2: Extract the original payload structure and business features, and combine them with the crawler verification mechanism to accurately identify user VIP level, resource preferences, social relationships (such as family / friends circle), and search browsing behavior, and build a VIP member rule base and resource ID recognition base.

[0048] Step 3: Introduce a self-learning optimization mechanism, integrating manual intervention with automatic hot patch updates to achieve full lifecycle management of member characteristics. The AI-driven feature recognition system elevates user behavior analysis capabilities to a new level, accurately identifying high-value users and supporting refined operations.

[0049] The specific implementation process of "OneID Account Graph":

[0050] Step 1: Integrate the content knowledge base and membership database to build a unified knowledge system across the entire network. This covers resources across all scenarios, including video, short video, music, reading, and e-commerce, achieving a 96% HTTP call detail record (CDR) recognition rate and supporting multi-dimensional parsing of apps, URLs, and keywords.

[0051] Step 2: By strongly linking mobile phone numbers with internet accounts, a network-wide account graph centered on OneID is formed, connecting multiple layers of relationship chains such as user groups, family circles, and friend circles.

[0052] Step 3: By dynamically connecting the crawler platform with the DPI engine, we can achieve integrated analysis of hundreds of millions of resource tags (shopping, search, microblog, etc.) to provide underlying capabilities for accurate user profiling and cross-business correlation analysis.

[0053] Example 2, based on Example 1, proposes a system for user value mining based on dynamic protocol parsing and multimodal AI, including the following modules:

[0054] Data acquisition and preliminary processing module: used for in-depth analysis of DPI raw bitstream data, accurately extracting core user access content; and combining intelligent algorithms to extract data features and behavioral patterns; in the data acquisition and preliminary processing module, web crawling is used to perform fusion verification and multi-dimensional information comparison to overcome the DPI recognition problem in complex network environments.

[0055] Dynamic Protocol Parsing Module: This module features enhanced DPI technology to automate the entire process of tracking protocol interactions in business applications. Based on fixed-mobile converged DPI deep parsing technology, it constructs a multi-layered protocol identification system, enabling full-process tracking and load content mining of mainstream app traffic characteristics. Through automated protocol interaction tracking, deep packet load detection, and XDR call detail record comparison analysis, it extracts key fields such as host, URL, and cookie. Combined with connection patterns and traffic statistics rules, it accurately identifies multiple protocol types and business scenarios. It also incorporates web crawling capabilities to capture public network content entities in real time based on URL resource attribute IDs, dynamically updating the parsing rule base to achieve dynamic translation capabilities for over 90% of resources, covering popular short videos and celebrity fan information, thus improving the real-time performance and coverage of content parsing.

[0056] The AI ​​intelligent recognition module leverages AI algorithms to perform deep learning on payload data, enabling multi-dimensional mining and secondary analysis of protocol features. It extracts the original payload structure and business characteristics, and, combined with a web crawler verification mechanism, accurately identifies user VIP levels, resource preferences, social relationships, and search and browsing behaviors, constructing a VIP member rule base and resource ID recognition base. A self-learning optimization mechanism is introduced, integrating manual intervention and automatic hot patch updates to achieve full lifecycle management of member features, accurately identifying high-value users and supporting refined operations. The AI ​​algorithm is a deep neural network algorithm, trained on a large amount of payload data to achieve multi-dimensional mining and secondary analysis of protocol features, improving the accuracy of user behavior and business feature recognition.

[0057] The OneID account graph construction module integrates a content knowledge base and a member database to build a unified knowledge system across the entire network, covering resources across all scenarios including video, short video, music, reading, and e-commerce. It achieves a 96% HTTP call detail record (CDR) recognition rate and supports multi-dimensional parsing of apps, URLs, and keywords. Through strong associations between mobile phone numbers and internet accounts, it forms a network-wide account graph centered on OneID, connecting multi-layered relationship chains such as user groups, family circles, and social circles. Combined with dynamic integration of a web crawler platform and the DPI engine, it enables the fusion analysis of hundreds of millions of resource tags, providing underlying capabilities for accurate user profiling and cross-business correlation analysis. Within the OneID account graph construction module, the integration of the content knowledge base and member database utilizes data fusion technology to uniformly process and store data from different sources and formats, constructing a unified knowledge system across the entire network.

[0058] The User Value Mining and Precision Service Module combines an innovative and optimized log analysis model to deeply mine user behavior patterns and preference characteristics across different platforms, improving the accuracy of key information identification. It constructs a dynamically updated user value assessment system, establishing personalized demand models through multi-dimensional data analysis to achieve precise benefit recommendations based on user profiles, improving service matching and commercial value conversion. Within this module, the dynamically updated user value assessment system is dynamically adjusted based on users' historical behavior data, real-time behavior data, and business scenario data to ensure the accuracy and timeliness of the assessment results.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A user value mining method based on dynamic protocol parsing and multimodal AI, characterized in that: Includes the following steps: Data collection and preliminary processing; Dynamic protocol parsing; AI intelligent recognition; OneID account graph construction; User Value Mining and Precision Service: By combining innovative and optimized log analysis models, we can deeply mine user behavior patterns and preference characteristics across different platforms to improve the accuracy of key information identification; we can build a dynamically updated user value assessment system, establish personalized demand models through multi-dimensional data analysis, and achieve precise benefit recommendations based on user profiles to improve service matching and commercial value conversion.

2. The user value mining method based on dynamic protocol parsing and multimodal AI according to claim 1, characterized in that: The specific operations for data collection and preliminary processing are as follows: Deeply analyze the raw DPI bitstream data to accurately extract the core content accessed by users; combine intelligent algorithms to extract data features and behavioral patterns.

3. The user value mining method based on dynamic protocol parsing and multimodal AI according to claim 2, characterized in that: The specific operations of dynamic protocol parsing are as follows: Build an automated, end-to-end DPI-enhanced technology to track the protocol interaction process of business applications; Based on the fixed-mobile fusion DPI deep analysis technology, a multi-level protocol identification system is constructed to realize the full-process tracking of traffic characteristics and load content mining of mainstream APPs; By automating protocol interaction tracking, deep packet load detection, and XDR call detail record comparison analysis, key fields such as host, URL, and cookie are extracted. Combined with connection patterns and traffic statistics rules, multiple protocol types and business scenarios are accurately identified. Web crawling capabilities are introduced to capture public network content entities in real time based on URL resource attribute IDs, dynamically update the parsing rule base, and achieve dynamic translation capabilities for more than 90% of resources, covering popular short videos and celebrity fan information, thereby improving the real-time performance and coverage of content parsing.

4. The user value mining method based on dynamic protocol parsing and multimodal AI according to claim 3, characterized in that: The specific operations of AI intelligent recognition are as follows: Relying on AI algorithms to perform deep learning on payload data, multi-dimensional mining and secondary analysis of protocol features are achieved; the original payload structure and business features are extracted, and combined with the crawler verification mechanism, user VIP level, resource preferences, social relationships and search browsing behavior are accurately identified to build a VIP member rule base and resource ID recognition base; a self-learning optimization mechanism is introduced, and manual intervention and automatic hot patch updates are integrated to achieve full lifecycle management of member features, accurately identify high-value users, and support refined operations.

5. The user value mining method based on dynamic protocol parsing and multimodal AI according to claim 4, characterized in that: The specific operations for building the OneID account graph are as follows: Integrate the content knowledge base and member database to build a unified knowledge system across the entire network, covering resources across all scenarios including video, short video, music, reading, and e-commerce, achieving a 96% HTTP call detail record recognition rate, and supporting multi-dimensional parsing of APP / URL / keywords; Through the strong association between mobile phone numbers and internet accounts, form a network-wide account graph centered on OneID, connecting multiple layers of relationship chains such as user groups, family circles, and friend circles; By dynamically connecting the web crawler platform with the DPI engine, we can achieve integrated analysis of hundreds of millions of resource tags, providing underlying capabilities to support accurate user profiling and cross-business correlation analysis.

6. A system for the user value mining method based on dynamic protocol parsing and multimodal AI as described in claim 5, characterized in that: Includes the following modules: Data acquisition and preliminary processing module: used for in-depth analysis of DPI raw bitstream data, accurate extraction of core user access content; and combined with intelligent algorithms to extract data features and behavioral patterns; Dynamic Protocol Parsing Module: It has the ability to build an automated, end-to-end DPI-enhanced technology to track the protocol interaction process of business applications; Based on the fixed-mobile fusion DPI deep analysis technology, a multi-level protocol identification system is constructed to realize the full-process tracking of traffic characteristics and load content mining of mainstream APPs; By automating protocol interaction tracking, deep packet load detection, and XDR call detail record comparison analysis, key fields such as host, URL, and cookie are extracted. Combined with connection patterns and traffic statistics rules, multiple protocol types and business scenarios are accurately identified. Web crawling capabilities are introduced to capture public network content entities in real time based on URL resource attribute IDs, dynamically update the parsing rule base, and achieve dynamic translation capabilities for more than 90% of resources, covering popular short videos and celebrity fan information, thereby improving the real-time performance and coverage of content parsing. AI Intelligent Recognition Module: Relying on AI algorithms to perform deep learning on payload data, it achieves multi-dimensional mining and secondary analysis of protocol features; it extracts the original payload structure and business features, and combines a web crawler verification mechanism to accurately identify user VIP levels, resource preferences, social relationships, and search and browsing behaviors, and builds a VIP member rule base and resource ID recognition base; it introduces a self-learning optimization mechanism, integrates manual intervention and automatic hot patch updates, realizes full lifecycle management of member features, accurately identifies high-value users, and supports refined operations; OneID Account Graph Construction Module: Integrates content knowledge base and member database to build a unified knowledge system across the entire network, covering resources across all scenarios including video, short video, music, reading, and e-commerce, achieving a 96% HTTP call detail record recognition rate, and supporting multi-dimensional parsing of APP / URL / keywords; through the strong association between mobile phone numbers and internet accounts, it forms a network-wide account graph centered on OneID, connecting multi-layered relationship chains of user groups, family circles, and friend circles; By combining the dynamic integration of the crawler platform and the DPI engine, the fusion analysis of hundreds of millions of resource tags is realized, providing underlying capability support for accurate user profiling and cross-business correlation analysis. User Value Mining and Precision Service Module: Combining innovative and optimized log parsing models, this module deeply mines user behavior patterns and preference characteristics across different platforms, improving the accuracy of key information identification; it constructs a dynamically updated user value assessment system, establishes personalized demand models through multi-dimensional data analysis, and achieves precise benefit recommendations based on user profiles, improving service matching and commercial value conversion.

7. The system according to claim 6, characterized in that: In the data acquisition and preliminary processing module, web crawling is used to perform fusion verification and multi-dimensional information comparison, overcoming the challenge of DPI identification in complex network environments.

8. The system according to claim 7, characterized in that: In the AI ​​intelligent recognition module, the AI ​​algorithm is a deep neural network algorithm. By training on a large amount of payload data, it realizes multi-dimensional mining and secondary analysis of protocol features, thereby improving the recognition accuracy of user behavior and business features.

9. A system according to claim 8, characterized in that: In the OneID account graph construction module, the content knowledge base and member database are integrated using data fusion technology to process and store data from different sources and in different formats in a unified manner, thereby building a unified knowledge system across the entire network.

10. A system according to claim 9, characterized in that: In the user value mining and precision service module, the dynamically updated user value assessment system is dynamically adjusted based on users' historical behavior data, real-time behavior data, and business scenario data to ensure the accuracy and timeliness of the assessment results.

Images