CN-122022866-A - AI-based full-automatic intelligent passenger acquisition system and method

Abstract

The invention relates to the technical field of artificial intelligence and digital marketing, in particular to a full-automatic intelligent passenger acquisition system and method based on AI, a data storage module, an AI intelligent processing module, a business analysis processing module, a function application module, a data interface integration module and a verification and auxiliary tracing module; according to the invention, by constructing an intelligent decision core consisting of a simulation and prediction engine and a multi-target optimization engine, virtual simulation verification and multi-target automatic optimizing can be carried out on the generated initial strategy, the strategy effect can be predicted in a simulation environment and the pareto optimal solution under a plurality of conflict targets such as touch, conversion, cost and the like can be automatically searched before real resources are input, the limitation of human decision is effectively overcome, and the marketing decision is updated from an experience mode of a braille to a data-driven and globally optimal intelligent calculation mode, so that the scientificity of the strategy and the input-output ratio of a customer are obviously improved.

Inventors

• ZHANG YONGDONG

Assignees

• 企智芯(北京)科技有限公司

Dates

Publication Date

20260512

Application Date

20260130

Claims (20)

1. 1. The full-automatic intelligent passenger acquisition system based on the AI comprises a data storage module, an AI intelligent processing module, a business analysis processing module, a function application module, a data interface integration module and a verification and auxiliary traceability module, and is characterized in that: the data storage module is used for intensively storing original product data, generated content assets, user and operation data required by the operation of the system and structured marketing decision knowledge autonomously extracted by the system; The AI intelligent processing module provides a core AI model and algorithm capability for the system, and integrates a professional engine for automatic content generation, strategy simulation and prediction, multi-objective optimization and decision-making interpretability analysis; The business analysis processing module is used for driving a specific marketing business process, and comprises the steps of automatically analyzing product information, creating multi-mode marketing content and generating an initial intelligent marketing strategy scheme based on multi-source data; The function application module encapsulates a core guest-obtaining function facing to a user and comprises product and IP analysis, content creation and release management and cue full life cycle management, wherein the function application module further comprises a functional unit for marketing IP role management, digital person video automatic generation, multi-platform automatic release management and cue automatic access and quality evaluation distribution, and the tasks are arranged and monitored through a central scheduling unit; The data interface integration module is used for realizing the safe connection and data interaction with an external social media platform, a client relationship management system and a third-party data source; the verification and auxiliary traceability module is used as a system level of independent operation, provides a forced intelligent decision closed loop for an initial strategy scheme generated by the business analysis processing module, and the closed loop sequentially executes the following processes: Firstly, virtual simulation verification is carried out on a strategy, and then multi-objective automatic optimizing is carried out on a feasible strategy; Generating an interpretable traceability report for the key decision; Finally, knowledge is refined through periodic global evaluation, and continuous evolution of system strategy capability is driven.
2. 2. The AI-based fully automatic intelligent acquisition system of claim 1, wherein the data storage module comprises a product and market data unit, a content asset unit, a cue and conversion data unit, a model and knowledge graph unit, wherein: The product and market data unit adopts a distributed hybrid storage architecture and is used for storing multi-format original product documents, structured product features, bidding materials and industry trend data; The content asset unit is used for storing multi-version marketing documents, audio and video materials, digital person models and synthesized content; the cue and conversion data unit is used for storing the full channel cue and scoring, distributing, following and converting full life cycle data thereof; the model and knowledge graph unit adopts a graph database for storing product-user-scene association knowledge and structured marketing decision rules extracted by the verification and auxiliary traceability module.
3. 3. The AI-based fully automatic intelligent guest-obtaining system according to claim 1, wherein the AI intelligent processing module comprises a natural language processing engine unit, a multi-modal data fusion engine unit, a simulation and prediction engine unit, a multi-objective optimization engine unit, and an interpretable AI engine unit, wherein: the natural language processing engine unit is established based on a Transformer architecture and is used for product document analysis, marketing document generation and semantic understanding; the multi-modal data fusion engine unit is established based on an attention mechanism and is used for aligning and associating text, images, videos and user behavior data to generate unified multi-modal feature vectors; The simulation and prediction engine unit builds a virtual marketing environment integrating the user behavior simulator, the platform recommendation algorithm simulator and the competition environment simulator based on reinforcement learning, and is used for carrying out multi-round deduction on the marketing strategy and outputting a predicted value of the key performance index; the multi-target optimization engine unit integrates a Bayesian optimization algorithm and an evolutionary algorithm, and is used for automatically optimizing a strategy parameter space under a plurality of mutually conflicting marketing targets and searching a pareto optimal solution set; The interpretive AI engine unit employs a feature attribution method based on an attention mechanism or a SHAP method for analyzing and visualizing decision key influencing factors and weights thereof.
4. 4. The AI-based fully automatic intelligent guest-obtaining system according to claim 1, wherein the business analysis processing module comprises a solution feasibility verification processor unit, a policy optimizing and tracing processor unit, wherein: the solution feasibility verification processor unit is configured to: Invoking a natural language processing engine and a multi-mode data fusion engine of the AI intelligent processing module, analyzing product information and generating an initial marketing strategy scheme and multi-mode marketing content; The policy optimizing and tracing processor unit is configured to: And calling the multi-target optimization engine unit to perform multi-target automatic optimization on the verified strategy, and calling the interpretable AI engine unit to generate a tracing report for explaining decision logic and key influence factors.
5. 5. The AI-based fully automatic intelligent guest-obtaining system according to claim 1, wherein the functional application module comprises a product intelligent analysis unit, an IP role management unit, an intelligent document generation unit, a digital human video authoring unit, a data monitoring and policy optimization unit, a multi-platform publication management unit, a clue automatic access unit, and a central intelligent dispatch center unit, wherein: (1) The product intelligent analysis unit is internally provided with a dynamic market feasibility assessment subunit which is used for calling the simulation and prediction engine unit to carry out virtual verification on the preliminary marketing strategy and outputting a structured product analysis report containing feasibility scores; (2) The intelligent document generation unit is configured to synchronously generate a plurality of alternative documents, call the simulation and prediction engine unit to carry out pre-evaluation through the pre-evaluation subunit, and simultaneously combine the feature similarity analysis with the historical high-performance documents to screen high-quality document schemes; (3) The digital human video creation unit is used for generating the file or script into digital human video content assets; (4) The multi-platform release management unit is used for carrying out platform specification adaptation on the content assets and automatically releasing the content assets; (5) The clue automatic access unit is used for automatically collecting clues from multiple channels and scoring, distributing and tracking; (6) The central intelligent scheduling center unit is internally provided with an intelligent decision board for visually displaying the state of the whole process task, the verification progress, the traceability report and the evaluation discovery.
6. 6. The AI-based fully automatic intelligent guest-obtaining system according to claim 1, wherein the central intelligent dispatch center unit is further configured to: in the whole-flow task scheduling logic, a verification and optimizing link is forcedly embedded into a key node related to strategy adjustment, and after a scheme feasibility verification processor unit and a strategy optimizing and tracing processor unit of the business analysis processing module are required to complete processing and pass, related strategy schemes can be scheduled to an execution stage.
7. 7. The AI-based fully automatic intelligent guest-obtaining system of claim 1, wherein the data interface integration module comprises a social media platform API interface unit, a cue platform docking interface unit, a third party data interface unit, and a file upload interface unit, wherein: The social media platform API interface unit adopts an OAuth2.0 authorization protocol and is used for safely docking the main stream social media platform to realize account authorization management, automatic content release and multidimensional data acquisition; The thread platform docking interface unit supports an API and Webhook mode and is used for docking with enterprise WeChat, form tools and a CRM system to realize real-time synchronization of thread data; the third party data interface unit is used for acquiring industry trend data, bid data or platform hot spot data; The file uploading interface unit is used for receiving PDF, word, excel or the product file in the picture format.
8. 8. The AI-based fully automatic intelligent guest-obtaining system according to claim 1, wherein the verification and auxiliary traceability module comprises a decision traceability interpretation unit, a system evolution evaluation unit, wherein: the decision trace source interpretation unit is configured to: Invoking the interpretable AI engine unit at a key decision point, generating a structured human-readable decision traceability report, and displaying the report through an intelligent decision board of the central intelligent scheduling center unit; the phylogenetic evaluation unit, as a separate evolutionary layer, is configured to: And automatically starting global duplication according to a preset period, comparing the characteristic difference of high-performance and low-performance content clusters through characteristic extraction, cluster analysis and causal inference technology, refining a structured marketing decision rule, and driving the model and a knowledge graph unit to update related AI models.
9. 9. The AI-based fully automatic intelligent guest obtaining system according to claim 5, wherein the IP character management unit comprises a character creation subunit, a character positioning subunit, a character material library subunit, a character policy subunit, and a multi-role management subunit, wherein the character positioning subunit automatically generates personal characteristics, image characteristics, language styles, professional domain settings, and target audience matching information of the IP character based on the product analysis report.
10. 10. The AI-based fully automatic intelligent acquisition system of claim 5, wherein the intelligent document generation unit comprises a timing schedule subunit, a document planning subunit, an intelligent authoring subunit, a document diversity control subunit, a document quality assessment subunit, and a document version management subunit, wherein the timing schedule subunit supports fixed time triggers, periodic triggers, or event driven triggers; The intelligent creation subunit adopts a large language model based on a Transformer architecture, combines a product knowledge base, an industry corpus and a marketing case base to generate a case, and continuously updates a generation strategy through reinforcement learning and/or preference optimization mechanisms based on historical content expression data.
11. 11. The AI-based fully automatic intelligent acquisition system of claim 5, wherein the digital person video authoring unit comprises a digital person image library subunit, a digital person customization subunit, a script generation subunit, a speech synthesis subunit, a lip-sync subunit, a scene rendering subunit, a video synthesis subunit, and a special effects addition subunit, wherein: The script generation subunit converts the script into a script comprising a minute mirror and a time axis; The voice synthesis subunit converts the script into voice and supports multilingual synthesis, emotion expression control and speech speed intonation adjustment; the lip synchronization subunit realizes digital population and voice synchronization; the video synthesis subunit synthesizes the digital person, the dubbing, the caption and the background into a complete video and supports batch generation; The special effect adding subunit is used for adding video elements such as transition, special effect, sticker or subtitle dynamic effect and the like to the complete video.
12. 12. The AI-based fully automatic intelligent acquisition system of claim 11, wherein the digital person video authoring unit employs a deep learning-based speech synthesis technique, a computer vision-based lip synchronization technique, and a 3D rendering-based scene generation technique, and the digital person customization subunit employs a GAN network and/or diffusion model-based digital person image generation technique.
13. 13. The AI-based full-automatic intelligent guest-obtaining system according to claim 5, wherein the data monitoring and policy optimizing unit comprises a self account monitoring subunit, a bid account monitoring subunit, a payoff content identifying subunit, a prediction model subunit and an intelligent optimizing subunit, wherein the prediction model subunit is used for predicting potential performance and/or trend change of content, and the intelligent optimizing subunit is used for automatically optimizing case direction, release time, tag keywords, video style or IP personnel setting parameters based on indexes such as play quantity, interaction data, vermicelli growth, conversion rate and negative feedback.
14. 14. The AI-based fully automatic intelligent guest-obtaining system according to claim 13, wherein the intelligent optimization subunit employs a multi-arm slot machine algorithm and a reinforcement learning algorithm to achieve continuous automatic optimization in dynamic balance between exploring new strategies and utilizing known preferred strategies.
15. 15. The AI-based full-automatic intelligent guest-obtaining system according to claim 5, wherein the multi-platform publication management unit comprises an account management subunit, an account authorization subunit, a platform adaptation subunit, a publication scheduling subunit, an automatic publication subunit, a publication status tracking subunit, an exception handling subunit and a batch publication subunit, wherein the platform adaptation subunit is configured to adapt to video specifications, title words, label rules and cover requirements, the publication scheduling subunit is configured to recommend an optimal publication time window and perform publication queue management and conflict detection, and the batch publication subunit is configured to perform batch publication of the same content asset to multiple platforms and/or multiple accounts.
16. 16. The AI-based fully automatic intelligent guest-obtaining system according to claim 5, wherein the thread automatic access unit comprises a thread acquisition subunit, a thread parsing subunit, a thread deduplication subunit, a thread classification subunit, a thread scoring subunit, a thread allocation subunit, and a thread tracking subunit, wherein the thread acquisition subunit is configured to acquire thread data comprising at least private messages, comments, forms, and stay components.
17. 17. The AI-based fully automatic intelligent acquisition system of claim 16, wherein the cue scoring subunit generates a cue quality score based on cue source weight, user interaction depth, purchase intent strength, user representation matching and historical conversion data, and the system calculates the acquisition ROI based on the cue conversion result and feeds back for policy optimization.
18. 18. The AI-based fully automatic intelligent guest-obtaining system according to claim 5, wherein the central intelligent dispatch center unit includes a resource allocation subunit, a flow monitoring subunit, an anomaly alert subunit and a data synchronization subunit for allocating computing and storage resources, monitoring task status, performing retries/alerts on failed tasks and ensuring data consistency among the modules.
19. 19. The AI-based fully automatic intelligent acquisition system of claim 1, further comprising a safety protection module, an intelligent learning module, and a report analysis module, wherein: The security protection module comprises account security protection, data encryption storage, access right control and an anticreeper mechanism; The intelligent learning module is used for continuously learning based on user feedback and/or effect data and triggering automatic model update; The report analysis module is used for generating a report with a guest obtaining effect, a report with a content representation, a report with an ROI analysis or a report with a trend prediction.
20. 20. AI-based full-automatic intelligent acquisition method using the AI-based full-automatic intelligent acquisition system as set forth in any one of claims 1 to 8, characterized by comprising the steps of: The method comprises the steps of firstly, collecting and structuring storage of multi-source data, namely collecting product documents, market data, user behavior data, bid information and historical marketing content through a data interface integration module, and storing the product documents, the market data, the user behavior data, the bid information and the historical marketing content into a data storage module; calling a natural language processing engine unit and a multi-mode data fusion engine unit of the AI intelligent processing module through a business analysis processing module, analyzing product characteristics and generating uniform multi-mode characteristic representation, and combining a knowledge graph and a rule base to generate an initial marketing strategy and multi-mode marketing content containing multiple sets of alternative schemes; The strategy virtual simulation and feasibility verification comprises the steps of calling a simulation and prediction engine unit of an AI intelligent processing module, inputting an initial marketing strategy into a virtual marketing environment constructed based on reinforcement learning to carry out multi-round deduction and A/B test, and outputting a key performance index predicted value and risk assessment; Step four, policy multi-objective optimizing and deciding, namely, invoking a multi-objective optimizing engine unit of an AI intelligent processing module for the verified policy, and automatically optimizing under a plurality of mutually conflicting marketing targets to obtain a pareto optimal policy combination; calling an interpretable AI engine unit of the AI intelligent processing module to perform feature attribution analysis on the decision process of the optimal strategy combination to generate a structured decision tracing report; step six, strategy scheduling execution and full-flow monitoring, namely scheduling execution of the strategy subjected to verification in the step three and optimization in the step four through a central intelligent scheduling center unit of the function application module, and management of content release and clue follow-up; and step seven, periodic global evaluation and system evolution, namely starting a system evolution evaluation unit of the verification and auxiliary traceability module according to a preset period, carrying out feature extraction and cluster analysis on the total marketing data, extracting a new structured marketing decision rule, updating the new structured marketing decision rule to a model and knowledge graph unit, driving related AI model parameter optimization, and realizing system self-adaptive evolution.

Description

AI-based full-automatic intelligent passenger acquisition system and method Technical Field The invention relates to the technical field of artificial intelligence and digital marketing, in particular to a full-automatic intelligent passenger acquisition system and method based on AI. Background With the rapid development of mobile internet and social media, short video, live broadcast and other content forms have become core channels for enterprises to acquire clients. Digital marketing is transitioning from rough delivery to sophisticated operations driven with data and intelligence. Artificial intelligence techniques, particularly natural language processing, computer vision and deep learning, are being widely used in marketing links to improve efficiency and effectiveness. The technical scheme is that a trans-former model-based cross-modal feature fusion model and a 'large consumption demand prediction-credit adaptation evaluation' double-prediction model are constructed by fusing client text, image, behavior and time sequence multi-modal data, so that the technical problems of single data dimension, demand and qualification prejudging section and insufficient strategy dynamics in traditional acquisition are solved, the potential client accurate identification and credit risk pre-management and control are realized, and the acquisition efficiency, client conversion quality and risk management and control capability of the large consumption scene are improved. The current enterprise digital marketing guest-obtaining system has three core faults, wherein tool faults cause the cutting of each link, the operation is complex and the cost is high, the intelligent faults are characterized by content dependent manual work, lack of simulation verification and automatic optimization of strategies and poor decision interpretation, closed loop faults cause marketing and sales disconnection, broken thread conversion funnels and ROIs to be difficult to measure, and the system cannot realize autonomous evolution through data driving and is difficult to adapt to rapid market change. In view of the foregoing, there is a need to design an AI-based fully automatic intelligent guest-obtaining system and method for solving the above-mentioned problems. Disclosure of Invention The invention aims to provide an AI-based full-automatic intelligent passenger acquisition system and method, so as to solve the defects in the prior art. In order to achieve the above object, the present invention provides the following technical solutions: the full-automatic intelligent passenger acquisition system based on the AI comprises a data storage module, an AI intelligent processing module, a business analysis processing module, a function application module, a data interface integration module and a verification and auxiliary tracing module; the data storage module is used for intensively storing original product data, generated content assets, user and operation data required by the operation of the system and structured marketing decision knowledge autonomously extracted by the system; The data storage module comprises a product and market data unit, a content asset unit, a clue and conversion data unit, a model and knowledge graph unit, wherein: The product and market data unit adopts a distributed hybrid storage architecture and is used for storing multi-format original product documents, structured product features, bidding materials and industry trend data; The content asset unit is used for storing multi-version marketing documents, audio and video materials, digital person models and synthesized content; the cue and conversion data unit is used for storing the full channel cue and scoring, distributing, following and converting full life cycle data thereof; the model and knowledge graph unit adopts a graph database for storing product-user-scene association knowledge and structured marketing decision rules extracted by the verification and auxiliary tracing module; The marketing decision rule stored in the model and knowledge graph unit exists in the form of "if (feature condition) then (policy suggestion), and the rule base is periodically refined and updated by the system evolution evaluation unit. The AI intelligent processing module provides a core AI model and algorithm capability for the system, and integrates a professional engine for automatic content generation, strategy simulation and prediction, multi-objective optimization and decision-making interpretability analysis; the AI intelligent processing module comprises a natural language processing engine unit, a multi-mode data fusion engine unit, a simulation and prediction engine unit, a multi-target optimization engine unit and an interpretable AI engine unit, wherein: the natural language processing engine unit is established based on a Transformer architecture and is used for product document analysis, marketing document generation and semantic understanding; the multi