Economic index visual arrangement method based on large-model multi-agent cooperation

Abstract

The invention discloses an economic index visual arrangement method based on large-model multi-agent collaboration, and relates to the technical field of multi-agent collaboration, and the method comprises the steps: processing received input information through a first large language model, generating a standardized demand, and decomposing the standardized demand into a plurality of subtasks; through a product manager agent, the plurality of subtasks are allocated to other agents in a preset agent role pool, and the agent role pool comprises a data analysis and development agent and a visual designer agent; determining a target data source through the data analysis and development agent and the first subtask allocated to the data analysis and development agent, and generating a data analysis result according to data in the target data source; and generating visual output corresponding to the data analysis result through the visual designer intelligent agent and the second subtask allocated to the visual designer intelligent agent. According to the method and the device, a completely autonomous process from demand understanding to final delivery is realized, so that the visualization efficiency of economic indexes is improved.

Description

Technical Field

[0001] The present application relates to the field of multi-agent collaboration technology, and in particular to a method for visualizing and arranging economic indicators based on large-scale multi-agent collaboration. Background Art

[0002] With the development of artificial intelligence technology, AI (Artificial Intelligence)-assisted development tools have initially achieved automatic processing and visualization of structured data through technologies such as natural language processing and machine learning.

[0003] Currently, AI-assisted tools for economic indicator analysis still require manual input of detailed parameters, rule definition, and the provision of specific algorithmic models to analyze and forecast economic indicators. They are unable to autonomously collect data, analyze, and interpret visualization results. This means they cannot achieve a fully autonomous process from demand understanding to final delivery. The entire economic indicator visualization process requires multiple manual interventions, resulting in low visualization efficiency.

[0004] The above content is only used to assist in understanding the technical solution of this application and does not constitute an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide an economic indicator visualization orchestration method based on large-scale model multi-agent collaboration, aiming to solve the technical problem of how to improve the efficiency of economic indicator visualization.

[0006] To achieve the above objectives, this application proposes a method for visualizing economic indicators based on large-scale model multi-agent collaboration, which includes: Processing received input information through the first language model to generate normalized requirements, and decomposing the normalized requirements into a plurality of subtasks, wherein the input information is user-inputted visualization requirement information for economic indicators; Allocate the multiple subtasks to other agents in a preset agent role pool through the product manager agent, wherein the agent role pool includes a data analysis and development agent and a visualization designer agent; Determine a target data source through the data analysis and development agent and the first subtask assigned thereto, and generate a data analysis result based on the data in the target data source; Generate a visualization output corresponding to the data analysis result through the visualization designer agent and its assigned second subtask.

[0007] In one embodiment, the step of determining the target data source through the data analysis and development agent and its assigned first subtask includes: Determining a target economic entity through the data analysis and development agent according to the first subtask; Determining, through the data analysis and development agent, a target matching degree between the target economic entity and each data source in a preset data source library; The target data source is determined by analyzing and developing the matching degree between the data and the target.

[0008] In one embodiment, the step of determining, by the data analysis and development agent, the target matching degree between the target economic entity and each data source in the preset data source library includes: Determine, by means of the data analysis and development agent, a standardized name of the target economic entity according to a preset economic knowledge graph; Determining, by the data analysis and development agent, a keyword matching degree between the standardized name and the indicator items of each of the data sources; Determining, by means of the data analysis and development agent, a semantic matching degree between the input information and the indicator items of each of the data sources; The target matching degree is determined according to the keyword matching degree and the semantic matching degree through the data analysis and development agent.

[0009] In one embodiment, the step of generating data analysis results based on the data in the target data source includes: Pulling target data from the target data source through the data analysis and development agent; The target data is cleaned by the data analysis and development agent to obtain cleaned target data; The data analysis and development agent retrieves the cleaned target data according to the first subtask to generate data analysis results.

[0010] In one embodiment, the step of searching the cleaned target data by the data analysis and development agent according to the first subtask to generate data analysis results includes: Determining a target economic entity through the data analysis and development agent according to the first subtask; Determining extended information of the target economic entity based on a preset economic knowledge graph; Performing a keyword search on the cleaned target data according to the extended information to obtain a first search result; Converting the input information and the extended information into text vectors using a fine-tuned large language model, wherein the fine-tuned large language model is obtained by fine-tuning training data in the economic field; Determining a second search result based on the vector similarity between the text vector and each indicator item in the cleaned target data; A target search result is determined based on the first search result and the second search result, and a data analysis result is generated based on the target search result.

[0011] In one embodiment, the agent role pool includes: a data architect agent, and after the step of pulling target data from the target data source, further includes: In the case where there are multiple target data sources, the data architect agent determines whether the indicator dimensions of the target data pulled from each target data source are consistent; In the case where the indicator dimensions of the target data are inconsistent, determining the target dimension through the data architect agent and the first subtask, and uniformly converting the indicator dimensions of the target data into the target dimension; Through the data architect agent, the target data are integrated.

[0012] In one embodiment, the agent role pool includes: a quality assurance agent, and after the step of generating data analysis results, further includes: Determining, by the quality assurance agent, whether a prediction deviation between a predicted value in the data analysis result and an obtained actual value exceeds a preset threshold; When the prediction deviation exceeds the preset threshold, the agent parameters of the data analysis and development agent are adjusted through a preset reinforcement learning strategy.

[0013] In one embodiment, the method further comprises: The input information and the output information of all agents in the agent role pool are stored uniformly to ensure that context is shared among all agents in the agent role pool.

[0014] In one embodiment, the step of processing the received input information using the first large language model to generate normalization requirements includes: Processing the input information through the large language model to determine a user intent and a semantic slot corresponding to the user intent; The semantic slots are filled according to the input information to obtain the normalization requirements.

[0015] In one embodiment, before the step of processing the received input information by the first large language model to generate normalization requirements, the method further includes: Performing intent recognition on the preset training data using the second language model to obtain a soft target for the training data; The third largest language model is trained using the training data and the soft target to obtain the first largest language model.

[0016] In addition, to achieve the above objectives, this application also proposes an economic indicator visualization orchestration system based on large-scale model multi-agent collaboration, which includes: A requirement parsing module, configured to process the received input information through the first language model, generate standardized requirements, and decompose the standardized requirements into multiple subtasks; A task assignment module is configured to assign the plurality of subtasks to other agents in a preset agent role pool through a product manager agent, wherein the agent role pool includes a data analysis and development agent and a visualization designer agent; a data analysis module, configured to determine a target data source through data analysis and development of the intelligent agent and its assigned first subtask, and generate data analysis results based on the data in the target data source; The data visualization module is used to generate a visualization output corresponding to the data analysis result through the visualization designer intelligent agent and the second subtask assigned to it.

[0017] In addition, to achieve the above-mentioned purpose, the present application also proposes an electronic device, which includes: a memory, a processor, and a computer program stored on the memory and runnable on the processor, and the computer program is configured to implement the steps of the economic indicator visualization orchestration method based on large-model multi-agent collaboration as described above.

[0018] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the economic indicator visualization orchestration method based on large-scale model multi-agent collaboration as described above are implemented.

[0019] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the economic indicator visualization orchestration method based on large-model multi-agent collaboration as described above.

[0020] One or more technical solutions proposed in this application have at least the following technical effects: first, receiving the visualization demand information for economic indicators input by the user, and processing the input information through the first language model to generate accurate and clear standardized requirements to reduce the deviation of the intelligent agent in understanding the user needs; then, through the first language model, decomposing the standardized requirements into multiple subtasks to reduce the time loss of manual task decomposition; then, through the product manager intelligent agent, assigning multiple subtasks to other intelligent agents in the preset intelligent agent role pool, providing a basis for realizing collaboration between multiple intelligent agents; then, through the data analysis and development intelligent agent in the intelligent agent role pool and its assigned first subtask, determining the target data source, and generating data analysis results based on the data in the target data source, avoiding the tedious operation of manually entering the data source to search and analyze data, and improving the efficiency and intelligence level of economic indicator analysis; then, through the visualization designer intelligent agent in the intelligent agent role pool and its assigned second subtask, converting the data analysis results into intuitive visualization output, ensuring that its visualization output meets user needs while reducing the operation of manually selecting visualization charts, and improving the efficiency of visualization result output. This application realizes a completely autonomous process from understanding economic indicator requirements to final delivery through collaboration between a large model and multiple intelligent agents. The large model and each intelligent agent can be automatically and seamlessly connected without human intervention, thereby improving the efficiency of economic indicator visualization in the overall operation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0023] Figure 1 This is a flowchart of the first embodiment of the method for visually arranging economic indicators based on large-scale multi-agent collaboration in this application; Figure 2 A schematic diagram of the process of visualizing economic indicators provided in Example 1 of this application; Figure 3 A schematic diagram of the process of economic data processing provided in Example 2 of this application; Figure 4 A schematic diagram of the multi-agent collaboration and communication process provided in Example 2 of this application; Figure 5This is the overall flow chart of multi-agent collaboration provided in Example 3 of this application; Figure 6 This is a schematic diagram of the module structure of the economic indicator visualization orchestration system based on large-scale model multi-agent collaboration in an embodiment of the present application; Figure 7 This is a schematic diagram of the device structure of the hardware operating environment involved in the method for visualizing economic indicator orchestration based on large-scale model multi-agent collaboration in an embodiment of the present application.

[0024] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0025] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0026] It should be noted that, in the description of this application specification and the appended claims, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0027] It should be noted that all actions of acquiring signals, information or data in this application are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0028] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0029] At present, the entire process of using AI-assisted tools to visualize economic indicators requires multiple manual interventions, resulting in low visualization efficiency.

[0030] This application realizes a completely autonomous process from understanding economic indicator requirements to final delivery through collaboration between a large model and multiple intelligent agents. The large model and each intelligent agent can be automatically and seamlessly connected without human intervention, thereby improving the efficiency of economic indicator visualization in the overall operation process.

[0031] It should be noted that the execution subject of this embodiment may be an electronic device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc.

[0032] Based on this, the embodiment of the present application provides a method for visualizing economic indicators based on large-scale model multi-agent collaboration, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the economic indicator visualization orchestration method based on large-scale model multi-agent collaboration in this application.

[0033] In this embodiment, the method for visualizing economic indicators based on large-scale model multi-agent collaboration includes steps S10 to S40: Step S10: Processing the received input information through the first language model to generate normalization requirements, and decomposing the normalization requirements into multiple subtasks; Input information is the user's visualization requirements for economic indicators, entered through a text box or voice input interface. It is presented in unstructured natural language and typically includes the name of the economic indicator, time range, and data presentation method (such as chart type). Economic indicators are quantitative characteristics used to measure the quantitative level or trend of economic activity, including GDP (Gross Domestic Product), CPI (Consumer Price Index), interest rates, etc.

[0034] A large language model (LLM) refers to a pre-trained deep learning model based on the Transformer architecture that has the ability to understand and generate natural language. The first large language model represents a large language model used to process user input information and generate standardized requirements.

[0035] Normalization requirements refer to the structured instruction set output by the first language model, which contains quantifiable parameters such as indicator type, time and space range, and can be presented in the form of JSON (JavaScript Object Notation).

[0036] Optionally, after receiving the input information entered by the user, it can be pre-processed by cleaning, word segmentation, part-of-speech tagging, etc., and potential entities in the input information can be identified, and the identified entities can be standardized, such as mapping the colloquial "gross domestic product" to a unified database identifier GDP; and then the standardized input information can be processed through the first language model.

[0037] Optionally, after a plurality of subtasks are generated, they and the standardized requirements may be further output to an interactive interface for user confirmation or adjustment.

[0038] Step S20: Allocate multiple subtasks to other agents in the preset agent role pool through the product manager agent; An agent is an artificially intelligent entity that can perceive its environment and take actions to achieve specific goals. The agent role pool is a cluster of pre-registered agents, each with independent task processing capabilities and a specific capability tag attached to it. This is used to match subtasks with user needs when assigning them.

[0039] The agent role pool includes product manager agents, data analysis and development agents, and visualization designer agents; among them, product manager agents refer to agents with task scheduling and resource allocation algorithms, which can realize sub-task allocation through capability label matching; data analysis and development agents refer to agents that integrate multiple data analysis algorithms and data processing tools, and are responsible for performing data-related processing tasks such as data collection, cleaning, conversion, feature engineering, model selection, algorithm implementation and data analysis; visualization designer agents refer to agents with built-in visualization design templates and graphics drawing tools to realize data visualization, and are responsible for designing visualization solutions according to data and sub-task requirements, including chart types, layouts, color schemes, etc.

[0040] For example, after the first language model converts the economic indicator visualization demand information input by the user into multiple subtasks, these subtasks will be sent to the product manager agent; then the product manager agent can match each subtask with the ability label of each agent in the agent role pool, and for any subtask, assign it to the agent with the highest matching score.

[0041] Optionally, multiple subtasks can be assigned to other agents in the preset agent role pool through the product manager agent and a preset scheduler, where the scheduler is used to manage the above information of the product manager agent and the output information of other agents.

[0042] For example, when the product manager agent is performing subtask assignment, long-term operation accumulation or sudden system failure may cause some of its data to be lost. At this time, the scheduler can obtain the above information to restore the lost task assignment data and ensure the continuity and accuracy of task assignment.

[0043] Optionally, the product manager agent determines the dependency relationships between multiple subtasks and assigns the multiple subtasks to other agents in the agent role pool; then, controls the other agents to execute their assigned subtasks according to the dependency relationships.

[0044] Step S30, determining the target data source through the data analysis and development agent and its assigned first subtask, and generating data analysis results based on the data in the target data source; The first subtask is used to represent the data processing-related subtasks assigned to the data analysis and development agent, which usually includes clear economic indicators, analysis requirements (such as growth rate) and time and space constraints.

[0045] The target data source refers to the data access point determined according to the analysis of the first subtask, which can be a database, an API (Application Programming Interface) provided by the official statistical structure, an unstructured document library, a real-time data stream, etc. This embodiment does not impose any specific restrictions on this.

[0046] Data analysis results refer to the information and conclusions obtained by the Data Analysis and Development Agent after searching, analyzing, and / or predicting the target data source according to the first subtask. Based on the first subtask, the Data Analysis and Development Agent can select an appropriate analysis model (for tasks such as analyzing relationships between economic indicators) or a prediction model (for tasks such as economic trend forecasting and risk assessment) to analyze or predict the retrieved data to generate data analysis results.

[0047] For example, the data analysis and development agent can analyze the economic indicators, time range and other information contained in the first subtask, search and match in a known data source list or database, and determine the target data source containing the required data; then, construct a data request instruction according to the first subtask, and send the instruction to the target data source to obtain a data set that meets the requirements of the first subtask; then, based on the data type and task requirements, determine the data analysis method and model, and use the determined data analysis method and model to process and analyze the retrieved data set to generate data analysis results.

[0048] In a feasible embodiment, the step of determining the target data source through the data analysis and development agent and its assigned first subtask in step S30 includes: Step S31, determining the target economic entity according to the first subtask through the data analysis and development agent; Economic entities refer to the objects of data association analysis, which can be economic indicators such as GDP and CPI, or microeconomic entities such as industries and enterprises, or economic events, economic policies, geographical regions, time dimensions, etc. This embodiment does not impose specific restrictions on this.

[0049] The target economic entity can be obtained by performing semantic analysis on the first subtask through the data analysis and development agent, which usually includes the indicator name, region and time.

[0050] Step S32: determining the target matching degree between the target economic entity and each data source in the preset data source library through data analysis and development of an intelligent agent; A data source library refers to a knowledge base that stores multiple data source information in a structured manner. Each entry records metadata such as the name of each data source, API interface, list of economic indicators covered by the data source (using standardized names), time granularity (year / quarter / month / day), geographic granularity (country / province / city), data update frequency, authority level of the data provider, and data quality level.

[0051] Target matching refers to the strength of the association between the target economic entity and each data source, which can be determined based on indicators such as semantic similarity, data dimension coverage and / or time granularity alignment.

[0052] In a feasible implementation, step S32 includes: Step S321: Determine the standardized name of the target economic entity based on the preset economic knowledge graph through data analysis and development of the intelligent agent; An economic knowledge graph is a graph-structured data structured by nodes (economic entities) and edges (relationships between entities). Node attributes include name, type, region, and business scope, while edge attributes include relationships such as affiliation, cooperation, and competition. A standardized name is a standard name that uniquely and regularly identifies an economic entity within the economic knowledge graph.

[0053] Optionally, before step S321, economic data can be obtained using open APIs provided by various economic data sources (including statistical yearbooks, financial reports, policy documents, research papers, news reports, industry analysis reports, etc.). The big model is then used to perform entity recognition, relationship extraction, and event extraction on this economic data to obtain economic entities (including economic events) and economic relationships in the economic field, and based on this, an economic knowledge graph is constructed. The big model is fine-tuned using economic text data annotated with entities, relationships, and events.

[0054] Optionally, new data can be dynamically acquired from the above economic data sources, the above knowledge extraction process can be repeated, and the economic knowledge graph can be dynamically updated and expanded based on the new knowledge extracted.

[0055] Optionally, during the dynamic construction of the economic knowledge graph, some knowledge graph content can be extracted through manual sampling to check whether there are problems such as entity recognition errors and inaccurate relationship extraction, so as to ensure the accuracy of the information in the economic knowledge graph.

[0056] Optionally, during the dynamic construction of the economic knowledge graph, extracted economic data can be analyzed and processed according to pre-set conflict detection and resolution rules. For example, when conflicting data extracted from different economic data sources arises (e.g., different economic data sources have different records of the same economic indicator), the more credible knowledge can be determined based on factors such as the authority of the economic data source (e.g., prioritizing data from official statistical departments) and the freshness of the data (e.g., selecting more recent data), and the economic knowledge graph can be updated accordingly.

[0057] It is understandable that by determining the standardized names of the target economic entities, the problem of inconsistent expressions of economic indicator names in different sources and scenarios can be effectively solved, which will help reduce errors and duplication of work caused by name differences and improve the efficiency of the entire economic indicator analysis and visualization process.

[0058] Step S322, determining the keyword matching degree between the standardized name and the index items of each data source through data analysis and development agent; Keyword matching refers to the degree of lexical overlap between the standardized name and the indicator items of the data source, which can be determined using the TF-IDF (Term Frequency-Inverse Document Frequency) algorithm, the BM25 (BestMatching 25) algorithm, and other algorithms.

[0059] Step S323, determining the semantic matching degree between the input information and the indicator items of each data source through data analysis and development of an intelligent agent; Semantic matching refers to the degree of matching obtained by comparing the deep meaning of the concepts expressed by the input information and the data source indicator items using natural language processing technology and semantic analysis algorithms. It is used to measure the consistency between the original indicator requirements and the data source indicators.

[0060] For example, for any data source, the BERT (Bidirectional Encoder Representations from Transformers) model can be used to convert the input information and the index items of the data source into vector form respectively, and the cosine similarity between the two vectors is calculated as the semantic matching degree between the input information and the index items of the data source.

[0061] Step S324: Through data analysis and development of intelligent agents, the target matching degree is determined based on the keyword matching degree and the semantic matching degree.

[0062] Illustratively, for any data source, a weighted sum calculation can be performed on the keyword matching degree and semantic matching degree between the target economic entity and the data source according to a preset weight distribution rule to obtain the target matching degree between the target economic entity and the data source.

[0063] It is understandable that the hybrid matching algorithm of keyword matching and semantic matching overcomes the limitations of relying solely on name matching or semantic matching, effectively avoiding the misselection of data sources due to similar names but inconsistent semantics, or similar semantics but large differences in names, thereby improving the accuracy and reliability of data acquisition and thus improving the accuracy of economic indicator analysis.

[0064] Step S33, determine the target data source through data analysis and development of the matching degree between the intelligent agent and each target.

[0065] For example, the data analysis and development agent may select a data source with the highest target matching degree or a data source that meets a certain matching degree threshold from the data source library as the target data source.

[0066] Optionally, when there are multiple data sources that can provide the same economic indicators, the priority of each data source that can provide the same economic indicators can be determined based on preset priority rules such as the data source's authority score, data update timeliness, data granularity matching, API stability, etc., and the data source with the highest priority can be determined as the target data source.

[0067] In this embodiment, the target data source is determined by combining the economic knowledge graph and the hybrid matching algorithm to improve the accuracy of data source matching and ensure that economic indicator data that meets user needs can be obtained.

[0068] Step S40: Generate a visualization output corresponding to the data analysis result through the visualization designer agent and its assigned second subtask.

[0069] The second subtask is used to represent the subtasks related to data visualization assigned to the visualization designer agent, which usually includes requirements in terms of visualization form, style, elements, etc.

[0070] Visual output refers to the data analysis results generated after processing by the visual designer intelligent agent and presented in an intuitive visual form, including interactive charts, analysis reports, large-screen components, etc.

[0071] For example, based on the second subtask, the Visual Designer agent can determine the horizontal and vertical axes and corresponding visualization formats, such as a line chart, pie chart, or stacked bar chart. Assuming a line chart is the visualization format, the Visual Designer agent can then use its built-in drawing tools and algorithms to create a line chart using data analysis of GDP changes over the past five years, with the year as the horizontal axis and the GDP value as the vertical axis, to reflect the changes in GDP values ​​over each year. Furthermore, the Visual Designer agent can automatically optimize the chart's layout, colors, fonts, labels, and other visual elements based on the potential audience and user preferences, ensuring the professionalism and readability of the visualization output.

[0072] For example, the visual designer agent and the second subtask assigned to it can generate a visual chart corresponding to the data analysis results, and automatically generate a chart title, legend and key insights based on the chart content to avoid rigid data piling; then the product manager agent can generate an introduction and background introduction based on the input information entered by the user; at the same time, the data analysis and development agent can generate key findings on economic indicators based on the data analysis methods and data analysis process it adopts, and generate conclusions and suggestions based on the data analysis results; then the product manager agent can generate an economic analysis report based on the above introduction, background introduction, data analysis methods, key findings, chart interpretation, conclusions and suggestions.

[0073] Optionally, before generating a visual output through the visual designer agent, a preset large model can be used to select a target chart from a preset chart library based on the data type of each data in the data analysis results and the user's visualization preference settings, and generate a preset number of candidate layout schemes that arrange the target chart; then, through the visual designer agent and the second subtask, the target layout scheme is determined from the candidate layout schemes, and the data analysis results are filled into the chart of the target layout scheme to obtain a visual output corresponding to the data analysis results.

[0074] Optionally, when the input information includes multiple economic indicators, the visualization designer agent can generate a corresponding visualization chart for each economic indicator; further, the visualization designer agent can arrange and layout the multiple generated visualization charts to generate a comprehensive interactive economic indicator dashboard, making it convenient for users to obtain information in one stop.

[0075] For example, please refer to Figure 2 , Figure 2A process diagram for visualizing economic indicators is provided. First, data input is provided to a visualization designer agent, including data analysis results and user visualization preferences. The visualization designer agent then performs visualization design based on the data input, during which it can provide a variety of visualization chart selections and layout schemes with the help of a large model driver. The visualization designer agent then determines the final chart composition and target layout scheme from the various layout schemes provided by the large model. The chart composition can be a line chart, a GIS (Geographic Information System) map, a multidimensional column chart, a composite chart, etc., and this embodiment does not impose specific restrictions on this. Data is then filled in according to the data analysis results to obtain a visualization chart, and a chart interpretation is generated according to the chart content and the data analysis results. Then, according to the presentation form in the second subtask, the visualization chart and the chart interpretation are combined to determine the final visualization output, wherein the presentation form can include a comprehensive dashboard, a data board, a comprehensive large screen, an analysis report, etc.

[0076] In this embodiment, through the collaboration between the big model and multiple intelligent agents, a completely autonomous process is achieved from understanding the economic indicator requirements to the final delivery. The big model and each intelligent agent can be automatically and seamlessly connected without human intervention, thereby improving the efficiency of economic indicator visualization in the overall operation process.

[0077] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction and will not be repeated hereafter. On this basis, the step of generating the data analysis result according to the data in the target data source in step S30 includes: Step S34, pulling target data from the target data source through the data analysis and development agent; Target data refers to the data set pulled from the target data source, which may include different data formats such as tables and JSON.

[0078] For example, the data analysis and development agent can select the appropriate protocol according to the type of the target data source, and construct an API request to send to the target data source; then receive the data stream returned by the target data source, and store it in a temporary cache area, waiting for the next step of processing.

[0079] In a feasible implementation, the agent role pool includes: a data architect agent, and after step S34, further includes: Step S341: When there are multiple target data sources, the data architect agent determines whether the indicator dimensions of the target data pulled from each target data source are consistent. A data architect agent is an agent that has the ability to analyze data structures and verify dimensions. It usually integrates algorithms such as data lineage analysis and data dimension alignment. It is responsible for designing data pipelines, determining data storage solutions, and planning data flows.

[0080] The indicator dimension refers to the set of metadata attributes of each field in the target data, including data type, unit, etc. For example, the "sales" field may be counted by "day / month" in different data sources.

[0081] Step S342: If the indicator dimensions of each target data are inconsistent, the target dimension is determined by the data architect agent and the first subtask, and the indicator dimensions of each target data are uniformly converted into the target dimension. The target dimension refers to the standard dimension for achieving unified analysis and comparison of data.

[0082] For example, when the data architect agent detects that the indicator dimensions of the target data are inconsistent, it analyzes the first subtask and determines the target dimensions that the user expects to display from the indicator dimensions corresponding to each target data according to the user's needs; then, the data architect agent uses professional conversion algorithms in the economic field (such as deflator, exchange rate conversion, etc.) to uniformly convert the indicator dimensions of each target data into target dimensions based on the dimensional differences.

[0083] Step S343: Integrate the target data through the data architect agent.

[0084] For example, the data architect agent can integrate the target data according to the standard name of each field in the target data. For example, according to the standard name, the data of the field within three years and three years ago can be integrated together for unified viewing and calling.

[0085] In this implementation, the data architect agent automatically performs dimension conversion and data integration to eliminate dimensional differences and data inconsistencies between different data sources, build a unified and comprehensive data set, provide a reliable data foundation for economic indicator analysis, and improve the accuracy of data analysis.

[0086] Step S35, performing data cleaning on the target data through the data analysis and development agent to obtain cleaned target data; Data cleaning refers to the process of detecting and correcting anomalies in target data based on economic rules or economic knowledge graphs, including data correction, outlier processing, missing value filling, and data quality verification.

[0087] For example, the data analysis and development agent can use a large language model to perform semantic understanding and contextual analysis of the target data, identify and correct typos and inconsistent formats in the target data; then use statistical methods and machine learning algorithms (such as the isolation forest algorithm, etc.) to automatically detect and process outliers; then intelligently fill in missing data based on time series forecasts, regression models or contextual information; and then combine the economic knowledge graph to perform logical consistency verification (for example, GDP cannot be less than CPI), data timeliness verification (ensuring that the data is the latest and available), and data integrity check.

[0088] It is understandable that data cleaning can effectively remove noise, errors and duplicate information in the data, improve the accuracy, completeness and consistency of the target data, and thus enhance the accuracy of economic indicator analysis.

[0089] In step S36, the data analysis and development agent searches the cleaned target data according to the first subtask to generate data analysis results.

[0090] Exemplarily, the data analysis and development agent parses the first subtask and determines the retrieval conditions, such as time range, field filtering rules, etc.; then, converts the retrieval conditions into the corresponding database query language or retrieval instructions, searches the cleaned target data, and obtains the retrieval results; then, according to the user needs in the first subtask, calls the relevant analysis algorithm or model, analyzes and processes the retrieval results, and generates data analysis results.

[0091] In a feasible implementation, step S36 includes: Step S361, determining the target economic entity through the data analysis and development agent according to the first subtask; The specific implementation of step S361 can refer to the specific implementation of step S31 in the first embodiment above, and will not be described in detail here.

[0092] Step S362: Determine the extended information of the target economic entity based on the preset economic knowledge graph; It should be noted that step S362 may adopt the economic knowledge graph constructed in the first embodiment.

[0093] Extended information refers to other economic concepts, entities or information related to the target economic entity obtained through economic knowledge graph reasoning.

[0094] For example, the data analysis and development agent can automatically expand the target economic entity to its synonyms, related entities, or hypernyms based on the entities and relationships in the economic knowledge graph, obtaining the corresponding extended information for the target economic entity to improve the search recall rate. For example, the entity "digital economy" can be expanded to "information industry," "big data," "artificial intelligence," etc.

[0095] Step S363: Perform keyword search on the cleaned target data based on the extended information to obtain a first search result; The first search result is used to represent the structured data obtained through keyword search.

[0096] Exemplarily, the data development and analysis agent can use a preset text retrieval algorithm (such as the BM25 algorithm) to perform keyword matching between different fields in the extended information and the indicator list of the cleaned target data to determine the matched target indicator; then, the data corresponding to the target indicator is extracted from the cleaned target data, and the first retrieval result can be obtained by combining the target indicator and the corresponding data.

[0097] It is understandable that by using keywords in the extended information to search the cleaned target data, we can further focus on the data related to the target economic entity to improve the data analysis effect.

[0098] Step S364: using the fine-tuned large language model, converting the input information and the extended information into text vectors, wherein the fine-tuned large language model is obtained by fine-tuning the training data in the economic field; The fine-tuned large language model is a specialized language model obtained by further training the original large language model on training data from the economic field. The text vector is a vector representation of the semantic features of the input and extended information extracted by the fine-tuned large language model.

[0099] Step S365, determining a second search result based on the vector similarity between the text vector and each index item in the cleaned target data; Vector similarity is a quantitative indicator used to measure the similarity between two vectors in high-dimensional space. It can be calculated using methods such as cosine similarity and Euclidean distance.

[0100] The second search result is used to represent the structured data determined by the vector similarity calculation.

[0101] Exemplarily, the data development and analysis agent can use the fine-tuned large language model to vectorize the input information and extended information to obtain a text vector; similarly, the fine-tuned large language model can be used to vectorize each indicator item in the cleaned target data to obtain multiple indicator vectors; and then the cosine similarity algorithm is used to calculate the vector similarity between the text vector and each indicator vector; and then the indicator item corresponding to the indicator vector whose vector similarity exceeds a preset threshold can be determined as the target indicator; and then the data corresponding to the target indicator is extracted from the cleaned target data, and the target indicator and the corresponding data are combined to obtain a second search result.

[0102] It can be understood that by calculating the vector similarity between the text vector and the target data index item, the second search result is determined, and accurate data retrieval based on semantic understanding is achieved, overcoming the problem that traditional keyword retrieval may miss semantically related but literally mismatched data.

[0103] Step S366: Determine a target search result based on the first search result and the second search result, and generate a data analysis result based on the target search result.

[0104] For example, the data development and analysis agent can combine the first search result and the second search result according to the indicator name, remove duplicate indicator items, and obtain the target search result; then, call the data analysis algorithm to analyze the target search result and generate a data analysis result.

[0105] Optionally, after obtaining the first and second search results, the indicators can be ranked based on factors such as their match with the input information, the authority of the corresponding data source, publication time, and connectivity within the economic knowledge graph (i.e., the strength of their association with other important economic concepts within the economic knowledge graph). The ranked indicators can then be combined to obtain the target search results. Subsequently, in the visualization process, indicators that meet user needs and have strong relevance can be prioritized for display, thereby enhancing the user's visualization experience.

[0106] For example, please refer to Figure 3 , Figure 3A flowchart for economic data processing is provided. The first step is to determine the data input, which can be read from the data source library. Each data source can be data from the National Bureau of Statistics, provincial data interfaces, industry reports, real-time financial data streams, etc.; then, data source discovery and connection are performed through the data analysis and development intelligent agent. After the target data source is determined, data is pulled from the target data source. When there are multiple target data sources, the target data from different data sources are dimensionally converted and integrated to obtain a complete target data (which can be presented in the form of data sets / data tables, etc.); then, the integrated target data is cleaned and verified, including data correction, outlier processing, missing value filling, and data quality verification, to obtain the cleaned target data; then, data retrieval is performed in combination with the knowledge graph, where the knowledge graph can be integrated from multiple knowledge graphs, such as an economic knowledge graph, a basic statistical knowledge graph, an industry private domain knowledge graph, etc.; data retrieval can include keyword retrieval and vector retrieval, and the target retrieval result obtained by combining the retrieval results of the two is the economic field data related to the economic indicators in the input information.

[0107] In this embodiment, by combining a hybrid search strategy of keyword search and vector search, the extensiveness of the data is ensured while the relevance and accuracy of the data are ensured, thereby improving the comprehensiveness and accuracy of the search results and thus improving the accuracy of economic indicator analysis.

[0108] In a feasible implementation, the agent role pool includes: a quality assurance agent, and after step S30, further includes: Step A10: Using the quality assurance agent, determine whether the prediction deviation between the predicted value in the data analysis result and the obtained actual value exceeds a preset threshold; A quality assurance agent refers to an agent with capabilities such as state perception, deviation calculation, root cause analysis, and model optimization. It is responsible for comprehensive testing and verification of data accuracy, model validity, visualization effects, and system functionality.

[0109] Optionally, during the entire life cycle of visual analysis of economic indicators, the entire visual orchestration process can be collaboratively monitored in real time through quality assurance agents, including: monitoring the connection status of the data source to ensure the stability of the data interface and uninterrupted data flow; monitoring the efficiency and resource consumption of the data cleaning, conversion, and integration processes to achieve data processing performance monitoring; tracking the running speed and accuracy of the prediction model or analysis algorithm; monitoring the efficiency of visualization generation to ensure that the generation speed of charts and reports meets the requirements; monitoring the usage of resources such as CPU (Central Processing Unit), memory, and storage, etc.

[0110] For example, the quality assurance agent can evaluate the data quality by monitoring the completeness, accuracy, consistency, timeliness, etc. of the target data.

[0111] For example, the quality assurance agent can evaluate the prediction model used by the data development and analysis agent during the data analysis process using metrics such as R-squared (R-squared), MAE (Mean Absolute Error), RMSE (Root Mean Squared Error), and MAPE (Mean Absolute Percentage Error). The analytical model used by the data development and analysis agent during the data analysis process can also be evaluated using metrics such as model goodness of fit, significance level, and residuals. The backpropagation algorithm and the evaluation results can then be used to adjust the prediction model or analytical model, or to replace the model used by the data development and analysis agent.

[0112] Step A20: When the prediction deviation exceeds a preset threshold, the agent parameters of the data analysis and development agent are adjusted through a preset reinforcement learning strategy.

[0113] Prediction bias refers to the absolute or relative error between the model's predicted value and the true value.

[0114] Reinforcement learning strategy is a machine learning method that updates the agent's behavior strategy (agent parameters) based on feedback reward signals through the interaction between the agent and the environment. It includes the PPO (Proximal Policy Optimization) algorithm and the SAC (Soft Actor-Critic) algorithm.

[0115] Among them, the agent parameters define the internal variables of the data analysis and development agent behavior and decision logic, which may include the called model, model parameters, feature weights, time decay factors, etc.

[0116] For example, when the data analysis and development agent calls a prediction model to generate data analysis results, the quality assurance agent periodically compares the predicted value in the data analysis result with the currently acquired time to determine the prediction deviation; and when the prediction deviation exceeds a preset threshold, a penalty signal is generated based on the prediction deviation; and then the agent parameters of the data analysis and development agent can be adjusted through the PPO algorithm and the penalty signal, such as adjusting the model parameters or the called prediction model.

[0117] In this embodiment, the agent parameters are automatically adjusted through reinforcement learning strategies, so that the data analysis and development agent automatically optimizes its own behavior when faced with excessive prediction deviations, gradually improving the accuracy and reliability of the prediction, reducing manual intervention, ensuring the efficiency of economic indicator analysis, and improving the reliability of economic indicator analysis.

[0118] In a feasible embodiment, the method further includes: In step A30, the input information and the output information of all agents in the agent role pool are uniformly stored to ensure that the context is shared among all agents in the agent role pool.

[0119] For example, any agent in the agent role pool can be composed of an independent large language model instance. When executing the corresponding assigned subtask, independent analysis can be performed within its large language model; and when the subtask is completed, the corresponding output information will be saved to a preset shared database so that other agents can retrieve the data.

[0120] Optionally, for any agent in the agent role pool, when a conflict is detected (such as lack of file creation permission), the timer starts, and communication is carried out with other agents in the agent role pool to resolve the conflict, such as asking the product manager agent whether it is possible not to create a new file; the timer stops when the conflict is resolved, but if the time when the conflict is detected exceeds a preset time threshold, an alarm message can be output to seek help from the user.

[0121] For example, please refer to Figure 4 , Figure 4A flow chart of multi-agent collaboration and communication is provided. First, the input information is processed through the first language model to determine the normalization requirements and multiple subtasks. The input information, normalization requirements and multiple subtasks are all stored in a unified shared memory. Then, the subtasks are assigned to different agents through the product manager agent, and the corresponding subtask assignment information is also stored in the shared memory. Then, each agent executes its assigned subtask. For example, the data analysis and development agent can determine the target data source according to its assigned subtask, and pull data for data retrieval and analysis. After pulling data from different data sources, it can use the data architect agent to perform data dimension conversion and data integration. The corresponding output information and communication context of the data analysis and development agent and the data architect agent are all stored in the shared memory for other agents to call. Then, the visualization designer agent can perform visualization design and chart analysis according to its assigned subtask and the data analysis results read from the shared memory, and output the final visualization results. During the entire life cycle of the visualization output, the quality assurance agent can monitor and comprehensively evaluate the execution process and data of other agents, and reversely adjust the agent parameters of each agent based on the evaluation results.

[0122] In this embodiment, by uniformly storing input information and agent output information, information flow between agents is achieved, duplication of work and information inconsistency are avoided, thereby ensuring the consistency of the entire analysis process and improving the efficiency of economic indicator visualization and the reliability of analysis results.

[0123] Based on the first and / or second embodiments of the present application, in the third embodiment of the present application, the same or similar contents as those in the first and second embodiments can be referred to above and will not be described in detail. On this basis, the step of processing the received input information by the first large language model in step S10 to generate the normalization requirements includes: Step S11: Process the input information through the first language model to determine the user intent and the semantic slot corresponding to the user intent; User intent refers to the core user request extracted by the large language model, and semantic slots refer to the key parameters to be filled under the user intent, which are determined autonomously by the large language model based on training data.

[0124] In a feasible implementation manner, before step S10, the method further includes: Step S01: performing intent recognition on preset training data using the second largest language model to obtain a soft target of the training data; Soft target refers to the probability distribution output by the teacher model during the knowledge distillation process, including the probability distribution of intent, entity type, and relationship type corresponding to each word or phrase in the training data, rather than a single category label.

[0125] The second large language model represents the teacher model used to output soft targets. It can be the original large language model, such as DeepSeek, Qwen, etc., or it can be a large language model obtained after fine-tuning with training data in the economic field. This embodiment does not impose specific restrictions on this.

[0126] Step S02: training the third language model using the training data and the soft target to obtain the first language model.

[0127] The third large language model represents the learning model in the knowledge distillation process, which is usually set to the original large language model.

[0128] For example, for a set of preset training data containing various economic analysis-related text samples, the second language model is used to identify the intent of these samples, and the probability distribution of each sample belonging to different intent categories is output to form a soft target; then, these training data and soft targets are input into the training framework, and the third language model is trained to learn how to generate an intent probability distribution similar to the soft target based on the input text, that is, to learn the "experience" of the teacher model (the second language model) in generalizing and processing fuzzy information; after multiple rounds of iterative optimization, the performance of the third language model reaches the expected level or the loss function converges, and the first language model is obtained to process actual user input.

[0129] For example, the first language model trained by distillation can be used to process input information. For the input information "I want to see the trends of China's GDP growth rate and CPI index in the past five years, preferably in a comparative chart", the corresponding identified user intent is "economic indicator trend comparison". The semantic slots under this user intent include economic indicators, regions, time ranges, visualization forms, etc.

[0130] In this embodiment, the first language model is trained by knowledge distillation, which can not only determine the economic entity from the input information, but also identify the relationship between different entities, thereby more accurately identifying the user intention corresponding to the input information.

[0131] Step S12: extract data content from the input information to fill the semantic slots and obtain normalization requirements.

[0132] Data content refers to the entity or attribute value related to the semantic slot in the input information.

[0133] For example, after extracting the data content, it can be directly filled in, or it can be filled in after inference using the first large language model. For example, for the input information "I want to see the trend of China's GDP growth rate and CPI index in the past five years," the data content corresponding to the region slot is determined to be "China," and it can be filled in directly. The data content corresponding to the time range slot is "the past five years," and the large language model can be used to determine the specific year information (such as 2020-2025) and then fill it in. Then, by combining the user intent and the filled semantic slots, the normalized requirements corresponding to the input information are obtained.

[0134] In this embodiment, the first language model is obtained through training using knowledge distillation technology, and intent recognition is performed using the first language model to generate standardized requirements, thereby improving the accuracy of user intent recognition and ensuring that the overall economic indicator visualization process is based on accurate user needs, reducing user intervention, and improving the efficiency of economic indicator visualization and the accuracy of economic indicator analysis results.

[0135] For example, in order to help understand the implementation process of the economic indicator visualization arrangement method based on large model multi-agent collaboration obtained by combining this embodiment with the above embodiment 1, please refer to Figure 5 , Figure 5The paper provides an overall flow chart of multi-agent collaboration. First, an interactive interface is provided, in which users can input their economic analysis requirements and visualization goals through the natural language text box or voice input interface. Then, the intelligent intention and demand analysis module analyzes the user input and outputs a preliminary standardized requirement and task decomposition plan to the interactive interface for user confirmation or fine-tuning. Then, the standardized requirement and multiple subtasks confirmed by the user are sent to the multi-agent collaborative management and orchestration module. The product manager agent is used to assign tasks, the data analysis and development agent is used to mine the target data source, perform data retrieval and data analysis, and the data architect agent is used to intelligently analyze the target data source. The multi-agent collaborative management and orchestration module transforms and integrates data from different data sources, selects and lays out visualization charts through a visual designer agent, and outputs final visualization results (including visual dashboards, interactive charts, and economic analysis reports). The multi-heterogeneous economic data processing and knowledge fusion module provides each agent with access to heterogeneous data sources and an economic knowledge graph, enabling them to obtain comprehensive and accurate economic indicator data for analysis. Furthermore, the interactive interface displays real-time progress information for each agent, such as "Data Architect Agent Designing Data Pipeline" and "Data Analysis and Development Agent Processing Data and Building Models," to help users understand the current project progress. Furthermore, the multi-agent collaborative management and orchestration module provides a quality assurance agent to continuously monitor the performance of the visualization output process and the stability of the underlying data sources, and optimizes each agent based on user suggestions for improvements to published results. Furthermore, when agents in the multi-agent collaborative management and orchestration module detect persistent conflicts or lack consensus within the agent, they can request a human supervisor (user) to resolve the conflict or negotiate based on user suggestions. Finally, after the complete visualization output is displayed on the interactive interface, users can trigger the publishing control to export the visualization results into documents, pictures, etc. with one click, or deploy them to the internal data platform for others to view and interact with.

[0136] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the economic indicator visualization orchestration method based on large-scale model multi-agent collaboration of the present application. More forms of simple transformations based on this technical concept are all within the scope of protection of the present application.

[0137] It should be noted that the economic indicator visualization orchestration method based on large-scale model multi-intelligent collaboration proposed in this application is not only limited to the field of economic indicator visualization, but can also be widely used in various scenarios that require efficient and autonomous management of complex information processing and decision-making processes. For example, in the supply chain and logistics optimization system, the intelligent agent team can collaboratively analyze data from different links (such as production, transportation, inventory, sales), autonomously identify bottlenecks, and generate optimization plans and visualization reports; in the construction of smart cities, the intelligent agent team can autonomously analyze and visualize urban operation data (such as traffic flow, environmental indicators, public safety incidents, population density), assist city managers in understanding the city’s operation status, predict development trends, and assist in formulating urban planning and emergency response strategies; it can also be applied to scenarios such as scientific research project management and data analysis, financial risk management and investment analysis, intelligent marketing and customer relationship management, to improve management efficiency, accuracy and intelligence in each scenario. This application does not impose specific restrictions on specific application scenarios.

[0138] The present application also provides a system for visualizing economic indicators based on large-scale multi-agent collaboration. Figure 6 The economic indicator visualization orchestration system based on large-scale model multi-agent collaboration includes: The requirement parsing module 10 is used to process the received input information through the first language model, generate standardized requirements, and decompose the standardized requirements into multiple subtasks; The task assignment module 20 is used to assign multiple subtasks to other agents in a preset agent role pool through the product manager agent, wherein the agent role pool includes a data analysis and development agent and a visualization designer agent; The data analysis module 30 is used to determine the target data source through data analysis and development of the intelligent agent and its assigned first subtask, and generate data analysis results based on the data in the target data source; The data visualization module 40 is used to generate a visualization output corresponding to the data analysis result through the visualization designer agent and its assigned second subtask.

[0139] The economic indicator visualization orchestration system based on large-scale multi-agent collaboration provided in the embodiment of the present application adopts the economic indicator visualization orchestration method based on large-scale multi-agent collaboration in the above embodiment, which can solve the technical problem of how to improve the efficiency of economic indicator visualization. Compared with the existing technology, the beneficial effects of the economic indicator visualization orchestration system based on large-scale multi-agent collaboration provided in the present application are the same as the beneficial effects of the economic indicator visualization orchestration method based on large-scale multi-agent collaboration provided in the above embodiment, and the other technical features of the economic indicator visualization orchestration system based on large-scale multi-agent collaboration are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0140] An embodiment of the present application provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the economic indicator visualization orchestration method based on large-model multi-agent collaboration in the above-mentioned embodiment one.

[0141] Reference below Figure 7 , which shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present application. The electronic devices in the embodiments of the present application may include, but are not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), in-vehicle terminals (such as in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 7 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0142] like Figure 7As shown, the electronic device may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory 1002 or programs loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the electronic device. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems may be connected to the input / output interface 1006: an input device 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; a storage device 1003 including, for example, a magnetic tape or hard disk; and a communication device 1009. The communication device 1009 may allow the electronic device to communicate with other devices wirelessly or wired to exchange data. Although the figures show electronic devices with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have instead.

[0143] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a read-only memory 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are performed.

[0144] The electronic device provided in the embodiment of the present application adopts the economic indicator visualization arrangement method based on large-scale model multi-agent collaboration in the above embodiment, which can solve the technical problem of how to improve the efficiency of economic indicator visualization. Compared with the existing technology, the beneficial effects of the electronic device provided in this application are the same as the beneficial effects of the economic indicator visualization arrangement method based on large-scale model multi-agent collaboration provided in the above embodiment, and the other technical features of the electronic device are the same as those disclosed in the method of the previous embodiment, which will not be repeated here.

[0145] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0146] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0147] An embodiment of the present application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, and the computer-readable program instructions are used to execute the economic indicator visualization orchestration method based on large-model multi-agent collaboration in the above-mentioned embodiment.

[0148] The computer-readable storage medium provided in the embodiments of the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0149] The computer-readable storage medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0150] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by an electronic device, the electronic device performs the functions defined in the method of the embodiment disclosed in this application.

[0151] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0152] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0153] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0154] The readable storage medium provided in the embodiments of this application is a computer-readable storage medium, storing computer-readable program instructions (i.e., a computer program) for executing the aforementioned method for visualizing economic indicators based on large-scale multi-agent collaboration. This computer-readable storage medium can address the technical problem of improving the efficiency of economic indicator visualization. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the method for visualizing economic indicators based on large-scale multi-agent collaboration provided in the aforementioned embodiments, and are not further elaborated here.

[0155] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned economic indicator visualization orchestration method based on large-model multi-agent collaboration.

[0156] The computer program product provided in the embodiments of this application can solve the technical problem of improving the efficiency of economic indicator visualization. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the economic indicator visualization orchestration method based on large-scale multi-agent collaboration provided in the above embodiments, and will not be elaborated here.

[0157] The above description is only part of the embodiments of the present application and does not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A method for visualizing economic indicators based on large-scale model multi-agent collaboration, characterized by: The method for visualizing economic indicators based on large-scale model multi-agent collaboration includes: Processing received input information through the first language model to generate normalized requirements, and decomposing the normalized requirements into a plurality of subtasks, wherein the input information is user-inputted visualization requirement information for economic indicators; Allocate the multiple subtasks to other agents in a preset agent role pool through the product manager agent, wherein the agent role pool includes a data analysis and development agent and a visualization designer agent; Determine a target data source through the data analysis and development agent and the first subtask assigned thereto, and generate a data analysis result based on the data in the target data source; Generate a visualization output corresponding to the data analysis result through the visualization designer agent and its assigned second subtask.

2. The method for visualizing economic indicators based on large-scale model multi-agent collaboration according to claim 1, characterized in that: The step of determining the target data source through the data analysis and development agent and the first subtask assigned thereto includes: Determining a target economic entity through the data analysis and development agent according to the first subtask; Determining, through the data analysis and development agent, a target matching degree between the target economic entity and each data source in a preset data source library; The target data source is determined by analyzing and developing the matching degree between the data and the target.

3. The method for visualizing economic indicators based on large-scale model multi-agent collaboration as claimed in claim 2 is characterized in that: The step of determining the target matching degree between the target economic entity and each data source in the preset data source library by the data analysis and development agent includes: Determine, by means of the data analysis and development agent, a standardized name of the target economic entity according to a preset economic knowledge graph; Determining, by the data analysis and development agent, a keyword matching degree between the standardized name and the indicator items of each of the data sources; Determining, by means of the data analysis and development agent, a semantic matching degree between the input information and the indicator items of each of the data sources; The target matching degree is determined according to the keyword matching degree and the semantic matching degree through the data analysis and development agent.

4. The method for visualizing economic indicators based on large-scale model multi-agent collaboration according to claim 1 is characterized in that: The step of generating data analysis results based on the data in the target data source includes: Pulling target data from the target data source through the data analysis and development agent; The target data is cleaned by the data analysis and development agent to obtain cleaned target data; The data analysis and development agent retrieves the cleaned target data according to the first subtask to generate data analysis results.

5. The method for visualizing economic indicators based on large-scale model multi-agent collaboration as claimed in claim 4 is characterized in that: The step of searching the cleaned target data by the data analysis and development agent according to the first subtask to generate data analysis results includes: Determining a target economic entity through the data analysis and development agent according to the first subtask; Determining extended information of the target economic entity based on a preset economic knowledge graph; Performing a keyword search on the cleaned target data according to the extended information to obtain a first search result; Converting the input information and the extended information into text vectors using a fine-tuned large language model, wherein the fine-tuned large language model is obtained by fine-tuning training data in the economic field; Determining a second search result based on the vector similarity between the text vector and each indicator item in the cleaned target data; A target search result is determined based on the first search result and the second search result, and a data analysis result is generated based on the target search result.

6. The method for visualizing economic indicators based on large-scale model multi-agent collaboration according to claim 4 is characterized in that: The agent role pool includes: a data architect agent, and after the step of pulling target data from the target data source, further includes: In the case where there are multiple target data sources, the data architect agent determines whether the indicator dimensions of the target data pulled from each target data source are consistent; In the case where the indicator dimensions of the target data are inconsistent, determining the target dimension through the data architect agent and the first subtask, and uniformly converting the indicator dimensions of the target data into the target dimension; Through the data architect agent, the target data are integrated.

7. The method for visualizing economic indicators based on large-scale model multi-agent collaboration according to claim 1 is characterized in that: The agent role pool includes: a quality assurance agent, and after the step of generating data analysis results, further includes: Determining, by the quality assurance agent, whether a prediction deviation between a predicted value in the data analysis result and an obtained actual value exceeds a preset threshold; When the prediction deviation exceeds the preset threshold, the agent parameters of the data analysis and development agent are adjusted through a preset reinforcement learning strategy.

8. The method for visualizing economic indicators based on large-scale model multi-agent collaboration according to claim 1 is characterized in that: The method further comprises: The input information and the output information of all agents in the agent role pool are stored uniformly to ensure that context is shared among all agents in the agent role pool.

9. The method for visualizing economic indicators based on large-scale model multi-agent collaboration according to claim 1, characterized in that: The step of processing the received input information by the first language model to generate normalization requirements includes: Processing the input information using the first language model to determine a user intent and a semantic slot corresponding to the user intent; The semantic slots are filled according to the input information to obtain the normalization requirements.

10. The method for visualizing economic indicators based on large-scale model multi-agent collaboration according to claim 9, characterized in that: Before the step of processing the received input information through the first large language model to generate normalization requirements, the method further includes: Performing intent recognition on the preset training data using the second language model to obtain a soft target for the training data; The third largest language model is trained using the training data and the soft target to obtain the first largest language model.

Images