A large-scale event operation optimization method and system based on multi-modal data and agents

By using multimodal data collection and intelligent agent collaborative scheduling, the problem of data silos in the operation and management system of large-scale events has been solved, enabling real-time decision-making and full-volume passenger flow aggregation, improving operational efficiency and member retention, and forming a complete operational closed loop.

CN123118145APending Publication Date: 2026-11-17BEIJING BEICHEN DIGITAL TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611334416.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-31
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

The large-scale event operation and management system suffers from severe data silos among its various business systems and a lack of unified governance, resulting in high operating costs and delayed decision-making. The management dashboard does not achieve multi-terminal adaptability, affecting the efficiency of real-time decision-making. The reliance on manual verification for free ticket holders means that passenger flow data cannot be fully collected, impacting the accuracy of congestion prediction. The lack of a lightweight passenger flow prediction model makes it impossible to predict risks in advance. Visitors lose contact once they leave the venue, and there is a lack of systematic operational support for member retention and repeat purchases, making it impossible to form a closed loop for passenger flow experience and long-term operation.

Method used

By collecting and standardizing multimodal data, and integrating intelligent agents such as passenger flow prediction, congestion warning, targeted advertising, business type recommendation, and member retention, real-time data connectivity and unified governance are achieved. Combined with lightweight time-series prediction models and digital verification and data entry modules, comprehensive strategy optimization solutions are generated to improve the intelligence and real-time nature of operational decisions.

Benefits of technology

It enables real-time data connectivity across the entire value chain, reduces operating costs, improves decision-making timeliness, provides a foundation of comprehensive customer flow data, predicts congestion risks in advance, generates differentiated membership operation strategies, and forms a complete operational closed loop from traffic acquisition to customer retention.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118145A_ABST
    Figure CN123118145A_ABST
Patent Text Reader

Abstract

The application relates to the technical field of wisdom tourism, in particular to a large-scale event operation optimization method and system based on multi-modal data and intelligent agents, which solves the problems that the existing optimization system has data islands, ticketing, passenger flow, consumption and other data are scattered, manual checking is required, decision lagging, the free ticket crowd relies on manual verification, passenger flow cannot be fully collected, and a passenger flow prediction model is lacked, risks cannot be predicted in advance, and member retention and repurchase lack systematic operation support, and the application comprises a data acquisition unit used for acquiring ticket reservation data, entry verification data, passenger flow trajectory data, consumption transaction data and user identity label data in the whole process of a large-scale event. Through multi-modal data acquisition and standardized collection and management, the application breaks through the data barriers among various business systems such as ticketing, passenger flow, consumption and portrait, forms a unified standard theme data warehouse, and enables full-link data to be real-time connected.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of smart cultural tourism technology, specifically to a method and system for optimizing the operation of large-scale events based on multimodal data and intelligent agents. Background Technology

[0002] With the rapid development of the digital economy, large-scale festivals and events (such as Spring Festival temple fairs, large-scale concerts, sports events, and exhibitions) are undergoing a profound transformation in their operation and management, shifting from extensive to intensive, digital, and intelligent approaches, serving as important carriers of urban cultural and tourism consumption. Currently, various large-scale events generally face complex challenges such as high concentrations of visitors, diverse consumer behaviors, numerous participants, and dispersed data sources, necessitating the use of digital means to achieve end-to-end data collection, fusion analysis, and intelligent decision-making. Simultaneously, the maturity of multimodal data fusion technology, multi-agent collaborative scheduling architecture, and large-scale model technology provides new technological pathways for the refined operation of large-scale events. How to effectively collect and manage multi-source heterogeneous data from ticketing systems, visitor flow monitoring, payment processing, and customer profiling, and how to dynamically optimize operational strategies based on data-driven approaches, has become a core issue in the digital upgrade of cultural and tourism festivals and events.

[0003] Existing large-scale event operation and management systems still have significant shortcomings: data silos are severe among various business systems, with ticketing, passenger flow, and consumption data scattered and lacking unified governance, requiring manual verification and correction, resulting in high operating costs and delayed decision-making; the management dashboard does not achieve multi-terminal adaptability, affecting real-time decision-making efficiency; the reliance on manual verification for free ticket holders leads to incomplete passenger flow data collection, affecting the accuracy of congestion prediction; the lack of lightweight passenger flow prediction models makes it impossible to predict risks in advance and guide resource scheduling; and the loss of contact with visitors upon departure results in a lack of systematic operational support for member retention and repeat purchases, failing to form a closed loop for passenger flow experience and long-term operation. Therefore, these systems do not meet current needs. To address this, we propose a large-scale event operation optimization method and system based on multimodal data and intelligent agents. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing the operation of large-scale events based on multimodal data and intelligent agents, in order to solve the significant defects of the large-scale event operation management system proposed in the background art: severe data silos between business systems, with ticketing, passenger flow, and consumption data scattered and lacking unified governance, requiring manual verification and correction, resulting in high operating costs and delayed decision-making; the management screen does not achieve multi-terminal adaptation, affecting the efficiency of real-time decision-making; the reliance on manual verification of free ticket holders leads to the inability to collect all passenger flow data, affecting the accuracy of congestion prediction; the lack of a lightweight passenger flow prediction model makes it impossible to predict risks in advance and guide resource scheduling; and the loss of contact with tourists upon departure, with a lack of systematic operational support for member retention and repeat purchases, failing to form a closed loop for passenger flow experience and long-term operation.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a large-scale event operation optimization system based on multimodal data and intelligent agents, comprising: The data acquisition unit is used to collect ticket reservation data, entry verification data, passenger flow trajectory data, consumption transaction data and user identity tag data throughout the entire process of large-scale events, and to label various types of data with timestamps, spatial coordinates and event scene attributes to form a multi-source heterogeneous raw data stream. The data collection and governance unit is used to receive raw data streams and operate according to preset standardized data interface specifications to form standardized datasets and store them in the subject data warehouse. The intelligent agent collaborative scheduling module is used to subscribe to real-time data streams from the theme data warehouse and is equipped with multiple independently operating and mutually communicating intelligent agents. The intelligent agents include at least a passenger flow prediction intelligent agent, a congestion warning intelligent agent, a precise promotion intelligent agent, a business type recommendation intelligent agent, and a member retention intelligent agent. The adaptive decision unit is used to receive real-time data and strategy suggestions from each intelligent agent and display the decision interface synchronously on the management screen, management backend and mobile terminal in a responsive layout. The adaptive decision unit includes a dynamic early warning module, which is used to display early warning information in the form of pop-ups, highlighted marks or push notifications. The digital verification and supplementary recording module is used to collect the admission ticket information and identity information of free-ticket or special groups, and convert them into standardized passenger flow records and write them into the subject data warehouse, so as to realize the digital supplementary recording of free-ticket groups and the collection of all passenger flow. The data security convergence gateway is used to establish an encrypted transmission channel between the government-side data platform and the enterprise-side operating system, and to realize the two-way secure exchange of government data and market operation data according to the preset permission management strategy and de-identification rules. The passenger flow prediction model engine is used to take historical event day passenger flow time series data, ticket pre-sale volume, weather data and holiday attributes as input features, output the passenger flow prediction volume and the heat map value of passenger flow density distribution in each area for the future preset period, and push them to the passenger flow prediction intelligent agent. The digital linkage platform is used to build an online activity space that includes an online reservation and tour guide portal and a live interactive module, and to transmit online user behavior data back to the data collection and governance unit. The digital interaction management module is used to uniformly configure and monitor the status of AR interactive devices and AI interactive terminals at the event site, collect user interaction behavior data and push it synchronously to the business recommendation intelligent agent and the member retention intelligent agent. The data collected by the data acquisition unit is standardized by the data collection and governance unit and then synchronously distributed to the intelligent agent collaborative scheduling module, the passenger flow prediction model engine, and the data security integration gateway. The intelligent agent collaborative scheduling module generates strategy optimization schemes based on prediction results and interactive data, and presents them through the adaptive decision-making unit.

[0006] Preferably, the congestion warning intelligent agent calculates the real-time congestion risk index of each area based on the predicted passenger flow density of each area provided by the passenger flow prediction intelligent agent, combined with the maximum safe carrying capacity of each area at the event site, according to the following formula: in Let be the congestion risk index for region i at time t. This is the predicted passenger flow density for region i at time t. The maximum safe carrying density threshold for region i; when When the congestion threshold is greater than the preset threshold, the congestion warning agent determines that region i is at risk of congestion and generates a congestion warning signal and corresponding traffic management suggestions.

[0007] Preferably, the member retention intelligent agent calculates the user's member value stratification index using an extreme value normalization accumulation method based on the user's visit frequency, cumulative consumption amount, average stay time and number of interaction participations. That is, it calculates the extreme value normalization value of each dimension indicator in all users and accumulates the normalization values ​​of each dimension. The member retention intelligent agent divides users into three levels based on the member value stratification index: high-value active users, medium-value potential users, and low-value churn risk users, and generates differentiated member operation strategies for each level.

[0008] Preferably, the business format recommendation agent calculates the real-time popularity index of each business format based on the number of consumer transactions, total consumer transaction amount, and related consumption frequency within the statistical window period, using the following formula: in This represents the real-time popularity index of business type j. To count the number of consumer transactions for business type j during the statistical window period, To calculate the total consumer transaction volume of business type j during the statistical window period, The set benchmark average order value, To count the number of related consumption transactions for business type j within the statistical window period.

[0009] Preferably, the business format recommendation agent also calculates the correlation consumption potential index between business format j and business format k according to the user overlap between different business formats, using the following formula: in The number of users who simultaneously consume in both business format j and business format k. Let j be the total number of users in the consumer business format. The total number of users in consumption format k; The value range of the related consumption potential index is 0 to 1. The higher the index value, the higher the overlap of the user groups of the two business formats, and the more suitable it is for cross-business format linkage configuration. When the associated consumption potential index exceeds a preset association threshold, the business format recommendation agent generates cross-business format joint marketing suggestions or business format proximity configuration suggestions.

[0010] Preferably, the lightweight time-series prediction model built into the passenger flow prediction model engine adopts a hybrid architecture combining a long short-term memory network and a seasonal autoregressive moving average model. It takes hourly passenger flow data, ticket pre-sale progress, daily weather forecast, and holiday type codes from the past several activity days as input features, and outputs the predicted passenger flow for each time period and the heat map value of passenger flow density distribution in each area. After each activity, the lightweight time-series prediction model incrementally learns and updates the model parameters using the actual passenger flow data of that activity.

[0011] Preferably, the precise advertising intelligent agent dynamically allocates the budget ratio of each advertising channel and outputs a channel-targeted delivery strategy based on the predicted passenger flow in each region provided by the passenger flow prediction intelligent agent and the customer profile tags in the standardized dataset, combined with the historical conversion rate data of each advertising channel. The precise advertising intelligent agent also generates multiple versions of creative content recommendation schemes based on the age distribution, geographical origin, and consumption preferences in the customer profile tags.

[0012] Preferably, when the data security convergence gateway exchanges data between the government-side data platform and the enterprise-side operating system, it performs the following steps: receiving a data request from the enterprise-side operating system containing the requester's identity identifier and the requested data range; determining whether the requester has access permissions for the corresponding data range based on the requester's identity identifier query permission management policy; if the requester has access permissions, dynamically de-identifying the requested data according to preset de-identification rules; establishing an encrypted transmission channel to send the de-identified data to the requester, and recording all operation information of this data exchange in the operation audit log.

[0013] An optimization method for a large-scale event operation optimization system using multimodal data and intelligent agents, the method comprising the following steps: S1: Through the data acquisition unit, real-time data collection is carried out on ticket reservation data, entry verification data, passenger flow trajectory data, consumption transaction data and user identity tag data throughout the entire process of large-scale events, and the data is labeled with timestamps, spatial coordinates and event scene attributes. S2: Input the multi-source heterogeneous raw data streams collected in step S1 into the data collection and governance unit, perform field mapping, format conversion, deduplication and cleaning and outlier removal according to the preset standardized data interface specifications, form a standardized dataset and store it in the theme data warehouse. S3: The digital verification and supplementary recording module collects entry voucher information and identity information when free-ticket or special groups enter the venue. It converts the information into standardized passenger flow records according to the standardized data interface specifications and writes them into the passenger flow data table of the theme data warehouse, thereby realizing the digital supplementary recording of free-ticket groups and the collection of all passenger flow. S4: Input the historical activity day passenger flow time series data, ticket pre-sale volume, weather data and holiday attributes from the theme data warehouse into the passenger flow prediction model engine. The passenger flow prediction model engine outputs the passenger flow prediction volume for the future preset period and the heat value of passenger flow density distribution in each area, and pushes the prediction results to the passenger flow prediction intelligent agent. S5: The intelligent agent collaborative scheduling module subscribes to real-time data streams from the topic data warehouse. Multiple intelligent agents on the intelligent agent collaborative scheduling module collaborate asynchronously and orchestrate tasks through message queues. The congestion warning intelligent agent determines whether there is a congestion risk based on the passenger flow prediction volume provided by the passenger flow prediction intelligent agent and the preset congestion threshold. When there is a congestion risk, it generates a congestion warning signal and diversion suggestions, which are synchronously displayed on each terminal through the adaptive decision unit. S6: The business format recommendation intelligent agent calculates the real-time popularity index and related consumption potential index of each business format based on the consumption transaction data and customer profile data in the standardized dataset, and combines the user preference data collected by the digital interaction management module, and generates dynamic business format recommendation schemes and merchant scheduling suggestions. S7: The member retention intelligent agent identifies high-frequency consumers and potential churned users based on the user identity tag data, consumption transaction data and interactive behavior data collected by the digital interaction control module in the standardized dataset, generates a tiered member operation strategy, and pushes member outreach suggestions and repurchase guidance plans to the operation personnel through the adaptive decision unit. S8: The precise promotion intelligent agent outputs targeted placement strategies and content creative suggestions for promotional channels based on the passenger flow prediction volume and customer profile data provided by the passenger flow prediction intelligent agent, and pushes them to the adaptive decision unit. S9: The intelligent agent collaborative scheduling module summarizes the output results of each intelligent agent to generate a comprehensive strategy optimization scheme. The comprehensive strategy optimization scheme includes at least a customer flow guidance strategy, a business format configuration strategy, a publicity and placement strategy, and a membership operation strategy. It is presented through an adaptive decision-making unit and pushes corresponding tour recommendations and activity information to the online activity space through a digital linkage platform to achieve simultaneous online traffic generation and offline experience.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention breaks down data barriers between various business systems such as ticketing, passenger flow, consumption, and profiling through multimodal data collection and standardized aggregation and governance, forming a unified standard thematic data warehouse. This enables real-time data flow across the entire chain without the need for manual verification and correction, significantly reducing operating costs and improving the timeliness of data-driven decision-making. 2. This invention, through the digital verification and supplementation module for free-admission groups, fully incorporates the entry data of free-admission and special groups into the total passenger flow statistics, upgrading the event passenger flow data from sampling estimation to full collection, providing a complete and accurate data foundation for passenger flow analysis and congestion prediction; 3. This invention uses a built-in lightweight time-series prediction model, taking historical passenger flow data, ticket pre-sales volume, weather and holiday attributes as input, to output the predicted passenger flow volume for future periods and the heat map value of passenger flow density distribution in each area. This enables operators to anticipate congestion risks in advance and formulate resource scheduling and precise promotion plans accordingly, realizing the transformation from passive response to proactive prediction. 4. This invention uses a multi-agent collaborative scheduling engine to carry multiple agents such as passenger flow prediction, congestion warning, precise publicity, business type recommendation and member retention. The agents collaborate asynchronously and arrange tasks through message queues, automatically generating a comprehensive strategy optimization plan that includes passenger flow guidance, business type configuration, publicity placement and member operation. This upgrades the operation decision-making of large-scale events from relying on human experience to data-driven intelligent collaborative decision-making. 5. This invention uses a member retention intelligent agent to comprehensively analyze multi-dimensional data such as user visit frequency, consumption amount, stay time and interaction behavior, divides users into different value levels and generates differentiated member operation strategies for each level, effectively solving the problems of tourists losing contact after leaving the venue and low member repurchase rate in large-scale event scenarios, forming a complete operation closed loop from traffic introduction to retention. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the large-scale event operation optimization method based on multimodal data and intelligent agents according to the present invention.

[0016] Figure 2 This is a schematic diagram of the internal agent collaboration relationship of the multi-agent collaborative scheduling engine of the present invention.

[0017] Figure 3 This is a schematic diagram of the data exchange process of the government and enterprise data security convergence gateway of the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0019] Please see Figures 1 to 3 One embodiment of the present invention provides a large-scale event operation optimization system based on multimodal data and intelligent agents, comprising: The data acquisition unit is used to collect ticket reservation data, entry verification data, passenger flow trajectory data, consumption transaction data and user identity tag data throughout the entire process of large-scale events, and to label various types of data with timestamps, spatial coordinates and event scene attributes to form a multi-source heterogeneous raw data stream. The data collection and governance unit is used to receive raw data streams and operate according to preset standardized data interface specifications to form standardized datasets and store them in the subject data warehouse. The intelligent agent collaborative scheduling module is used to subscribe to real-time data streams from the theme data warehouse and is equipped with multiple independently operating and mutually communicating intelligent agents. The intelligent agents include at least a passenger flow prediction intelligent agent, a congestion warning intelligent agent, a precise promotion intelligent agent, a business type recommendation intelligent agent, and a member retention intelligent agent. The adaptive decision unit is used to receive real-time data and strategy suggestions from each intelligent agent and display the decision interface synchronously on the management screen, management backend and mobile terminal in a responsive layout. The adaptive decision unit includes a dynamic early warning module, which is used to display early warning information in the form of pop-ups, highlighted marks or push notifications. The digital verification and supplementary recording module is used to collect the admission ticket information and identity information of free-ticket or special groups, and convert them into standardized passenger flow records and write them into the subject data warehouse, so as to realize the digital supplementary recording of free-ticket groups and the collection of all passenger flow. The data security convergence gateway is used to establish an encrypted transmission channel between the government-side data platform and the enterprise-side operating system, and to realize the two-way secure exchange of government data and market operation data according to the preset permission management strategy and de-identification rules. The passenger flow prediction model engine is used to take historical event day passenger flow time series data, ticket pre-sale volume, weather data and holiday attributes as input features, output the passenger flow prediction volume and the heat map value of passenger flow density distribution in each area for the future preset period, and push them to the passenger flow prediction intelligent agent. The digital linkage platform is used to build an online activity space that includes an online reservation and tour guide portal and a live interactive module, and to transmit online user behavior data back to the data collection and governance unit. The digital interaction management module is used to uniformly configure and monitor the status of AR interactive devices and AI interactive terminals at the event site, collect user interaction behavior data and push it synchronously to the business recommendation intelligent agent and the member retention intelligent agent. The data collected by the data acquisition unit is standardized by the data collection and governance unit and then synchronously distributed to the intelligent agent collaborative scheduling module, the passenger flow prediction model engine, and the data security integration gateway. The intelligent agent collaborative scheduling module generates strategy optimization schemes based on prediction results and interactive data, and presents them through the adaptive decision-making unit.

[0020] Preferably, the congestion warning intelligent agent calculates the real-time congestion risk index of each area based on the predicted passenger flow density of each area provided by the passenger flow prediction intelligent agent, combined with the maximum safe carrying capacity of each area at the event site, according to the following formula: in Let be the congestion risk index for region i at time t. This is the predicted passenger flow density for region i at time t. The maximum safe carrying density threshold for region i; when When the congestion threshold is greater than the preset threshold, the congestion warning agent determines that region i is at risk of congestion and generates a congestion warning signal and corresponding traffic management suggestions.

[0021] Preferably, the member retention intelligent agent calculates the user's member value stratification index using an extreme value normalization accumulation method based on the user's visit frequency, cumulative consumption amount, average stay time and number of interaction participations. That is, it calculates the extreme value normalization value of each dimension indicator in all users and accumulates the normalization values ​​of each dimension. The member retention intelligent agent divides users into three levels based on the member value stratification index: high-value active users, medium-value potential users, and low-value churn risk users, and generates differentiated member operation strategies for each level.

[0022] Preferably, the business format recommendation agent calculates the real-time popularity index of each business format based on the number of consumer transactions, total consumer transaction amount, and related consumption frequency within the statistical window period, using the following formula: in This represents the real-time popularity index of business type j. To count the number of consumer transactions for business type j during the statistical window period, To calculate the total consumer transaction volume of business type j during the statistical window period, The set benchmark average order value, To count the number of related consumption transactions for business type j within the statistical window period.

[0023] Preferably, the business format recommendation agent also calculates the correlation consumption potential index between business format j and business format k according to the user overlap between different business formats, using the following formula: in The number of users who simultaneously consume in both business format j and business format k. Let j be the total number of users in the consumer business format. The total number of users in consumption format k; The value range of the related consumption potential index is 0 to 1. The higher the index value, the higher the overlap of the user groups of the two business formats, and the more suitable it is for cross-business format linkage configuration. When the associated consumption potential index exceeds a preset association threshold, the business format recommendation agent generates cross-business format joint marketing suggestions or business format proximity configuration suggestions.

[0024] Preferably, the lightweight time-series prediction model built into the passenger flow prediction model engine adopts a hybrid architecture combining a long short-term memory network and a seasonal autoregressive moving average model. It takes hourly passenger flow data, ticket pre-sale progress, daily weather forecast, and holiday type codes from the past several activity days as input features, and outputs the predicted passenger flow for each time period and the heat map value of passenger flow density distribution in each area. After each activity, the lightweight time-series prediction model incrementally learns and updates the model parameters using the actual passenger flow data of that activity.

[0025] Preferably, the precise advertising intelligent agent dynamically allocates the budget ratio of each advertising channel and outputs a channel-targeted delivery strategy based on the predicted passenger flow in each region provided by the passenger flow prediction intelligent agent and the customer profile tags in the standardized dataset, combined with the historical conversion rate data of each advertising channel. The precise advertising intelligent agent also generates multiple versions of creative content recommendation schemes based on the age distribution, geographical origin, and consumption preferences in the customer profile tags.

[0026] Preferably, when the data security convergence gateway exchanges data between the government-side data platform and the enterprise-side operating system, it performs the following steps: receiving a data request from the enterprise-side operating system containing the requester's identity identifier and the requested data range; determining whether the requester has access permissions for the corresponding data range based on the requester's identity identifier query permission management policy; if the requester has access permissions, dynamically de-identifying the requested data according to preset de-identification rules; establishing an encrypted transmission channel to send the de-identified data to the requester, and recording all operation information of this data exchange in the operation audit log.

[0027] An optimization method for a large-scale event operation optimization system using multimodal data and intelligent agents, the method comprising the following steps: S1: Through the data acquisition unit, real-time data collection is carried out on ticket reservation data, entry verification data, passenger flow trajectory data, consumption transaction data and user identity tag data throughout the entire process of large-scale events, and the data is labeled with timestamps, spatial coordinates and event scene attributes. S2: Input the multi-source heterogeneous raw data streams collected in step S1 into the data collection and governance unit, perform field mapping, format conversion, deduplication and cleaning and outlier removal according to the preset standardized data interface specifications, form a standardized dataset and store it in the theme data warehouse. S3: The digital verification and supplementary recording module collects entry voucher information and identity information when free-ticket or special groups enter the venue. It converts the information into standardized passenger flow records according to the standardized data interface specifications and writes them into the passenger flow data table of the theme data warehouse, thereby realizing the digital supplementary recording of free-ticket groups and the collection of all passenger flow. S4: Input the historical activity day passenger flow time series data, ticket pre-sale volume, weather data and holiday attributes from the theme data warehouse into the passenger flow prediction model engine. The passenger flow prediction model engine outputs the passenger flow prediction volume for the future preset period and the heat value of passenger flow density distribution in each area, and pushes the prediction results to the passenger flow prediction intelligent agent. S5: The intelligent agent collaborative scheduling module subscribes to real-time data streams from the topic data warehouse. Multiple intelligent agents on the intelligent agent collaborative scheduling module collaborate asynchronously and orchestrate tasks through message queues. The congestion warning intelligent agent determines whether there is a congestion risk based on the passenger flow prediction volume provided by the passenger flow prediction intelligent agent and the preset congestion threshold. When there is a congestion risk, it generates a congestion warning signal and diversion suggestions, which are synchronously displayed on each terminal through the adaptive decision unit. S6: The business format recommendation intelligent agent calculates the real-time popularity index and related consumption potential index of each business format based on the consumption transaction data and customer profile data in the standardized dataset, and combines the user preference data collected by the digital interaction management module, and generates dynamic business format recommendation schemes and merchant scheduling suggestions. S7: The member retention intelligent agent identifies high-frequency consumers and potential churned users based on the user identity tag data, consumption transaction data and interactive behavior data collected by the digital interaction control module in the standardized dataset, generates a tiered member operation strategy, and pushes member outreach suggestions and repurchase guidance plans to the operation personnel through the adaptive decision unit. S8: The precise promotion intelligent agent outputs targeted placement strategies and content creative suggestions for promotional channels based on the passenger flow prediction volume and customer profile data provided by the passenger flow prediction intelligent agent, and pushes them to the adaptive decision unit. S9: The intelligent agent collaborative scheduling module summarizes the output results of each intelligent agent to generate a comprehensive strategy optimization scheme. The comprehensive strategy optimization scheme includes at least a customer flow guidance strategy, a business format configuration strategy, a publicity and placement strategy, and a membership operation strategy. It is presented through an adaptive decision-making unit and pushes corresponding tour recommendations and activity information to the online activity space through a digital linkage platform to achieve simultaneous online traffic generation and offline experience.

[0028] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A large event operation optimization system based on multi-modal data and agents, characterized in that, include: The data acquisition unit is used to collect ticket reservation data, entry verification data, passenger flow trajectory data, consumption transaction data and user identity tag data throughout the entire process of large-scale events, and to label various types of data with timestamps, spatial coordinates and event scene attributes to form a multi-source heterogeneous raw data stream. The data collection and governance unit is used to receive raw data streams and operate according to preset standardized data interface specifications to form standardized datasets and store them in the subject data warehouse. The intelligent agent collaborative scheduling module is used to subscribe to real-time data streams from the theme data warehouse and is equipped with multiple independently operating and mutually communicating intelligent agents. The intelligent agents include at least a passenger flow prediction intelligent agent, a congestion warning intelligent agent, a precise promotion intelligent agent, a business type recommendation intelligent agent, and a member retention intelligent agent. The adaptive decision unit is used to receive real-time data and strategy suggestions from each intelligent agent and display the decision interface synchronously on the management screen, management backend and mobile terminal in a responsive layout. The adaptive decision unit includes a dynamic early warning module, which is used to display early warning information in the form of pop-ups, highlighted marks or push notifications. The digital verification and supplementary recording module is used to collect the admission ticket information and identity information of free-ticket or special groups, and convert them into standardized passenger flow records and write them into the subject data warehouse, so as to realize the digital supplementary recording of free-ticket groups and the collection of all passenger flow. The data security convergence gateway is used to establish an encrypted transmission channel between the government-side data platform and the enterprise-side operating system, and to realize the two-way secure exchange of government data and market operation data according to the preset permission management strategy and de-identification rules. The passenger flow prediction model engine is used to take historical event day passenger flow time series data, ticket pre-sale volume, weather data and holiday attributes as input features, output the passenger flow prediction volume and the heat map value of passenger flow density distribution in each area for the future preset period, and push them to the passenger flow prediction intelligent agent. The digital linkage platform is used to build an online activity space that includes an online reservation and tour guide portal and a live interactive module, and to transmit online user behavior data back to the data collection and governance unit. The digital interaction management module is used to uniformly configure and monitor the status of AR interactive devices and AI interactive terminals at the event site, collect user interaction behavior data and push it synchronously to the business recommendation intelligent agent and the member retention intelligent agent. The data collected by the data acquisition unit is standardized by the data collection and governance unit and then synchronously distributed to the intelligent agent collaborative scheduling module, the passenger flow prediction model engine, and the data security integration gateway. The intelligent agent collaborative scheduling module generates strategy optimization schemes based on prediction results and interactive data, and presents them through the adaptive decision-making unit.

2. The multi-modal data and agent based large event operations optimization system of claim 1, wherein, The congestion early warning intelligent agent calculates the real-time congestion risk index of each area based on the predicted passenger flow density of each area provided by the passenger flow prediction intelligent agent, combined with the maximum safe carrying capacity of each area at the event site, according to the following formula: in Let be the congestion risk index for region i at time t. This is the predicted passenger flow density for region i at time t. The maximum safe carrying density threshold for region i; When When the congestion risk exists in the region i, the congestion early warning agent generates a congestion early warning signal and a corresponding congestion early warning signal.

3. The multi-modal data and agent based large event operations optimization system of claim 1, wherein, The member retention intelligent agent calculates the user's member value stratification index based on the user's visit frequency, cumulative consumption amount, average stay time and number of interaction participations using the extreme value normalization accumulation method. That is, it calculates the extreme value normalization value of each dimension indicator in all users and accumulates the normalization values ​​of each dimension. The member retention intelligent agent divides users into three levels based on the member value stratification index: high-value active users, medium-value potential users, and low-value churn risk users, and generates differentiated member operation strategies for each level.

4. The large-scale event operation optimization system based on multimodal data and intelligent agents according to claim 1, characterized in that, The business type recommendation agent calculates the real-time popularity index for each business type based on the number of consumer transactions, total consumer transaction amount, and related consumption frequency within the statistical window period, using the following formula: in This represents the real-time popularity index of business type j. To count the number of consumer transactions for business type j during the statistical window period, To calculate the total consumer transaction volume of business type j during the statistical window period, The set benchmark average order value, To count the number of related consumption transactions for business type j within the statistical window period.

5. The multi-modal data and agent based large event operations optimization system of claim 4, wherein, The business format recommendation agent also calculates the correlation consumption potential index between business format j and business format k based on the user overlap between different business formats using the following formula: wherein is the number of users consuming both format j and format k, is the total number of users consuming format j, is the total number of users consuming format k; The value range of the related consumption potential index is 0 to 1. The higher the index value, the higher the overlap of the user groups of the two business formats, and the more suitable it is for cross-business format linkage configuration. When the associated consumption potential index exceeds a preset association threshold, the business format recommendation agent generates cross-business format joint marketing suggestions or business format proximity configuration suggestions.

6. The multi-modal data and agent based large event operations optimization system of claim 1, wherein, The lightweight time-series prediction model built into the passenger flow prediction model engine adopts a hybrid architecture combining a long short-term memory network and a seasonal autoregressive moving average model. It takes hourly passenger flow data from several past event days, ticket pre-sale progress, daily weather forecast, and holiday type codes as input features, and outputs the predicted passenger flow for each time period and the heat map value of passenger flow density distribution in each area. After each event, the lightweight time-series prediction model incrementally learns and updates the model parameters using the actual passenger flow data of that event.

7. The multi-modal data and agent based large event operations optimization system of claim 1, wherein, The precise advertising intelligence agent dynamically allocates the budget ratio of each advertising channel based on the predicted passenger flow in each region provided by the passenger flow prediction intelligence agent and the customer profile tags in the standardized dataset, combined with the historical conversion rate data of each advertising channel, and outputs a channel-targeted delivery strategy through a multi-armed slot machine algorithm. At the same time, the precise advertising intelligence agent generates multiple versions of creative content recommendation schemes based on the age distribution, geographical origin and consumption preferences in the customer profile tags.

8. The multi-modal data and agent based large event operations optimization system of claim 1, wherein, When the data security convergence gateway exchanges data between the government-side data platform and the enterprise-side operating system, it performs the following steps: receiving a data request from the enterprise-side operating system containing the requester's identity identifier and the requested data range; determining whether the requester has access permissions for the corresponding data range based on the requester's identity identifier query permission management policy; if the requester has access permissions, dynamically de-identifying the requested data according to preset de-identification rules; establishing an encrypted transmission channel to send the de-identified data to the requester, and recording all operation information of this data exchange in the operation audit log.

9. The optimization method of the multi-modal data and agent's large-scale event operation optimization system according to any one of claims 1-8, characterized in that, The method includes the following steps: S1: Through the data acquisition unit, real-time data collection is carried out on ticket reservation data, entry verification data, passenger flow trajectory data, consumption transaction data and user identity tag data throughout the entire process of large-scale events, and the data is labeled with timestamps, spatial coordinates and event scene attributes. S2: Input the multi-source heterogeneous raw data streams collected in step S1 into the data collection and governance unit, perform field mapping, format conversion, deduplication and cleaning and outlier removal according to the preset standardized data interface specifications, form a standardized dataset and store it in the theme data warehouse. S3: The digital verification and supplementary recording module collects entry voucher information and identity information when free-ticket or special groups enter the venue. It converts the information into standardized passenger flow records according to the standardized data interface specifications and writes them into the passenger flow data table of the theme data warehouse, thereby realizing the digital supplementary recording of free-ticket groups and the collection of all passenger flow. S4: Input the historical activity day passenger flow time series data, ticket pre-sale volume, weather data and holiday attributes from the theme data warehouse into the passenger flow prediction model engine. The passenger flow prediction model engine outputs the passenger flow prediction volume for the future preset period and the heat value of passenger flow density distribution in each area, and pushes the prediction results to the passenger flow prediction intelligent agent. S5: The intelligent agent collaborative scheduling module subscribes to real-time data streams from the topic data warehouse. Multiple intelligent agents on the intelligent agent collaborative scheduling module collaborate asynchronously and orchestrate tasks through message queues. The congestion warning intelligent agent determines whether there is a congestion risk based on the passenger flow prediction volume provided by the passenger flow prediction intelligent agent and the preset congestion threshold. When there is a congestion risk, it generates a congestion warning signal and diversion suggestions, which are synchronously displayed on each terminal through the adaptive decision unit. S6: The business format recommendation intelligent agent calculates the real-time popularity index and related consumption potential index of each business format based on the consumption transaction data and customer profile data in the standardized dataset, and combines the user preference data collected by the digital interaction management module, and generates dynamic business format recommendation schemes and merchant scheduling suggestions. S7: The member retention intelligent agent identifies high-frequency consumers and potential churned users based on the user identity tag data, consumption transaction data and interactive behavior data collected by the digital interaction management module in the standardized dataset, generates a tiered member operation strategy, and pushes member outreach suggestions and repurchase guidance plans to the operation personnel through the adaptive decision unit. S8: The precise promotion intelligent agent outputs targeted placement strategies and content creative suggestions for promotional channels based on the passenger flow prediction volume and customer profile data provided by the passenger flow prediction intelligent agent, and pushes them to the adaptive decision unit. S9: The intelligent agent collaborative scheduling module summarizes the output results of each intelligent agent to generate a comprehensive strategy optimization scheme. The comprehensive strategy optimization scheme includes at least a customer flow guidance strategy, a business format configuration strategy, a publicity and placement strategy, and a membership operation strategy. It is presented through an adaptive decision-making unit and pushes corresponding tour recommendations and activity information to the online activity space through a digital linkage platform to achieve simultaneous online traffic generation and offline experience.