CN120746508B - A Visualization and Arrangement Method for Economic Indicators Based on Large-Model Multi-Agent Collaboration - Google Patents

Abstract

This application discloses a method for visualizing and orchestrating economic indicators based on a large-scale model and multi-agent collaboration, relating to the field of multi-agent collaboration technology. The method includes: processing received input information through a first large-scale language model to generate standardized requirements, and decomposing these requirements into multiple sub-tasks; assigning these sub-tasks to other agents in a pre-defined 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; determining the target data source through the data analysis and development agent and its assigned first sub-task, and generating data analysis results based on the data in the target data source; and generating the visualization output corresponding to the data analysis results through the visualization designer agent and its assigned second sub-task. This application achieves a fully autonomous process from requirement understanding to final delivery, thereby improving the visualization efficiency of economic indicators.

Description

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

Technical Field

The application relates to the technical field of multi-agent coordination, in particular to an economic index visualization arrangement method based on large-model multi-agent coordination.

Background

With the development of artificial intelligence technology, AI (ARTIFICIAL INTELLIGENCE ) auxiliary development tools preliminarily realize automatic processing and visual presentation of structural data through natural language processing, machine learning and other technologies.

At present, for economic index analysis tasks, an AI auxiliary tool still needs to manually input detailed parameters, define rules and provide a specific algorithm model to analyze and predict the economic index, and cannot autonomously perform links such as data collection, analysis and visual result interpretation, namely cannot realize a complete autonomous flow from demand understanding to final delivery. In the whole economic index visualization process, manual intervention is needed for many times, so that the visualization efficiency is low.

The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present application and is not intended to represent an admission that the foregoing is prior art.

Disclosure of Invention

The application mainly aims to provide an economic index visualization arrangement method based on large-model multi-agent cooperation, which aims to solve the technical problem of how to improve the efficiency of economic index visualization.

In order to achieve the above purpose, the application provides an economic index visualization arrangement method based on large-model multi-agent cooperation, which comprises the following steps:

processing received input information through a first large language model, generating a normalized demand, and decomposing the normalized demand into a plurality of subtasks, wherein the input information is visual demand information aiming at economic indexes and input by a user;

Distributing the subtasks to other agents in a preset agent role pool through a product manager agent, wherein the agent role pool comprises data analysis and development agents and visual designer agents;

Determining a target data source through data analysis and development of the intelligent agent and a first subtask distributed to the intelligent 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 subtasks distributed to the visual designer intelligent agent.

In one embodiment, the step of determining the target data source by data analysis and development of the agent and its assigned first subtask includes:

determining a target economic entity according to the first subtask through the data analysis and development agent;

determining target matching degree between the target economic entity and each data source in a preset data source library through the data analysis and development intelligent agent;

And determining a target data source through the data analysis and development agent and the matching degree of each target.

In one embodiment, the step of determining, by 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 includes:

Determining the standardized name of the target economic entity according to a preset economic knowledge graph through the data analysis and development intelligent agent;

determining the keyword matching degree between the standardized names and the index items of the data sources through the data analysis and development agent;

determining semantic matching degree between the input information and index items of each data source through the data analysis and development agent;

and determining the target matching degree according to the keyword matching degree and the semantic matching degree through the data analysis and development agent.

In one embodiment, the step of generating the data analysis result according to the data in the target data source includes:

Pulling target data from the target data source through the data analysis and development agent;

performing data cleaning on the target data through the data analysis and development intelligent agent to obtain cleaned target data;

And searching the cleaned target data according to the first subtask by the data analysis and development agent to generate a data analysis result.

In one embodiment, the step of retrieving the cleaned target data by the data analysis and development agent according to the first subtask, and generating a data analysis result includes:

determining a target economic entity according to the first subtask through the data analysis and development agent;

according to a preset economic knowledge graph, determining the expansion information of the target economic entity;

according to the expansion information, keyword retrieval is carried out on the cleaned target data, and a first retrieval result is obtained;

Converting the input information and the expansion information into text vectors by utilizing a trimmed large language model, wherein the trimmed large language model is obtained by trimming training data in the economic field;

determining a second retrieval result according to the vector similarity between the text vector and each index item in the cleaned target data;

And determining a target search result according to the first search result and the second search result, and generating a data analysis result according to the target search result.

In one embodiment, the agent role pool includes a data architect agent, after the step of pulling target data from the target data source, further comprising:

In the case that a plurality of target data sources exist, judging whether index dimensions of target data pulled from each target data source are consistent or not through the data architect agent;

under the condition that the index dimensions of the target data are inconsistent, determining target dimensions through the data architect agent and the first subtask, and uniformly converting the index dimensions of the target data into the target dimensions;

integrating each target data by the data architect agent.

In one embodiment, the agent role pool includes a quality assurance agent, and after the step of generating the data analysis result, further includes:

Judging whether the predicted deviation between the predicted value in the data analysis result and the obtained actual value exceeds a preset threshold value or not through the quality assurance agent;

and under the condition that the prediction deviation exceeds the preset threshold value, the data analysis and the intelligent agent parameter of the intelligent agent development are adjusted through a preset reinforcement learning strategy.

In an embodiment, the method further comprises:

And uniformly storing the input information and the output information of all the agents in the agent role pool so as to ensure that the context is shared among all the agents in the agent role pool.

In one embodiment, the step of generating normalized requirements by processing the received input information through the first large language model includes:

processing the input information through the large language model, and determining user intention and semantic slots corresponding to the user intention;

And filling the semantic slots according to the input information to obtain the normalized requirements.

In one embodiment, before the step of generating the normalized requirement by processing the received input information through the first large language model, the method further includes:

Performing intention recognition on preset training data through a second large language model to obtain a soft target of the training data;

And training a third large language model through the training data and the soft target to obtain the first large language model.

In addition, in order to achieve the above object, the present application further provides an economic indicator visual arrangement system based on large-model multi-agent cooperation, the economic indicator visual arrangement system based on large-model multi-agent cooperation comprising:

The demand analysis module is used for processing the received input information through the first large language model, generating a normalized demand and decomposing the normalized demand into a plurality of subtasks;

The task allocation module is used for allocating the subtasks to other agents in a preset agent role pool through product manager agents, wherein the agent role pool comprises data analysis and development agents and visual designer agents;

the data analysis module is used for determining a target data source through data analysis and development of the intelligent agent and the first subtasks distributed to the intelligent agent, and generating a data analysis result according to data in the target data source;

And the data visualization module is used for generating a visual output corresponding to the data analysis result through the visual designer intelligent agent and the second subtasks distributed to the designer intelligent agent.

In addition, in order to achieve the aim, the application also provides electronic equipment, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the computer program is configured to realize the steps of the economic index visualization arrangement method based on large model multi-agent cooperation.

In addition, in order to achieve the above object, the present application further provides a storage medium, which is a computer readable storage medium, and the storage medium stores a computer program, and the computer program when executed by a processor implements the steps of the economic index visualization arrangement method based on large-model multi-agent collaboration as described above.

In addition, to achieve the above object, the present application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the economic index visualization arrangement method based on large model multi-agent collaboration as described above.

The technical scheme provided by the application has at least the following technical effects that firstly, visual demand information aiming at economic indexes and input by a user is received, the input information is processed through a first large language model, accurate and clear standardization demands are generated, so that understanding deviation of the agents to the user demands is reduced, the standardization demands are further decomposed into a plurality of subtasks through the first large language model, so that time loss of task disassembly is reduced manually, further, the plurality of subtasks are distributed to other agents in a preset agent role pool through a product manager agent, a basis is provided for realizing cooperation among the plurality of agents, further, a target data source is determined through data analysis and development agent in the agent role pool and a first subtask distributed to the agent role pool, a data analysis result is generated according to data in the target data source, tedious operation of manually inputting the data source for searching and analyzing the data is avoided, the efficiency and the intelligent level of economic index analysis are improved, further, the visual analysis result is converted into visual output charts by the visual design agent in the agent role pool and a second subtask distributed to the visual agent role pool, visual operation efficiency of the visual analysis result is improved, visual output of the visual analysis result is ensured, and visual requirements of users are simultaneously met. According to the application, through cooperation between the large model and the multiple intelligent agents, a complete autonomous process from understanding of economic index requirements to final delivery is realized, and the large model and the intelligent agents can be automatically and seamlessly connected without manual intervention, so that the efficiency of economic index visualization is improved on the whole operation process.

Drawings

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

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the description of the embodiments or the prior art will be briefly described below, and it will be obvious to those skilled in the art that other drawings can be obtained from these drawings without inventive effort.

FIG. 1 is a schematic flow chart of an embodiment of a large-model multi-agent collaborative economic index visual arrangement method according to the present application;

FIG. 2 is a schematic flow chart of an economic indicator visualization according to a first embodiment of the present application;

FIG. 3 is a schematic flow chart of an economic data processing according to a second embodiment of the present application;

fig. 4 is a schematic flow chart of multi-agent collaboration and communication according to a second embodiment of the present application;

FIG. 5 is a general flow chart of multi-agent collaboration provided in accordance with a third embodiment of the present application;

FIG. 6 is a schematic diagram of a module structure of an economic indicator visualization arrangement system based on large-model multi-agent collaboration in an embodiment of the application;

Fig. 7 is a schematic diagram of a device structure of a hardware operating environment related to an economic indicator visual arrangement method based on large-model multi-agent cooperation in an embodiment of the application.

The achievement of the objects, functional features and advantages of the present application will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

It should be understood that the specific embodiments described herein are merely illustrative of the technical solution of the present application and are not intended to limit the present application.

It is noted that, in the description of the present specification and the appended claims, the terms "first", "second", and the like are used solely to distinguish the description and are not to be construed as indicating or implying relative importance.

It should be noted that, all actions of acquiring signals, information or data in the present application are performed under the condition of conforming to the corresponding data protection rule policy of the country of the location and obtaining the authorization given by the owner of the corresponding device.

For a better understanding of the technical solution of the present application, the following detailed description will be given with reference to the drawings and the specific embodiments.

At present, in the whole process of realizing the visualization of the economic index by using the AI auxiliary tool, manual intervention is needed for a plurality of times, so that the visualization efficiency is lower.

According to the application, through cooperation between the large model and the multiple intelligent agents, a complete autonomous process from understanding of economic index requirements to final delivery is realized, and the large model and the intelligent agents can be automatically and seamlessly connected without manual intervention, so that the efficiency of economic index visualization is improved on the whole operation process.

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

Based on this, the embodiment of the application provides a large-model multi-agent collaboration-based economic index visualization arrangement method, and referring to fig. 1, fig. 1 is a flow diagram of a first embodiment of the large-model multi-agent collaboration-based economic index visualization arrangement method.

In this embodiment, the large-model multi-agent collaboration-based economic index visualization arrangement method includes steps S10-S40:

step S10, processing the received input information through a first large language model, generating a normalized demand, and decomposing the normalized demand into a plurality of subtasks;

The input information is visual requirement information for the economic index, which is input by a user through a text box or a voice input interface, is presented in an unstructured natural language form, and generally comprises the contents of an economic index name, a time range, a data presentation mode (such as a chart type) and the like. And economic indicators refer to quantitative characteristics for measuring the quantity level or trend of change of economic activities, including GDP (Gross Domestic Product, domestic total production), CPI (consumer price index, consumption price index), interest rate, etc.

The large language model (Large Language Model, LLM) refers to a pre-training deep learning model based on a transducer architecture, and has natural language understanding and generating capability, and the first large language model characterizes the large language model for processing user input information and generating normalized requirements.

The normalized requirement refers to a structured instruction set output by the first large language model, which contains quantifiable parameters, such as index type, space-time range, and the like, and can be presented in the form of JSON (JavaScript Object Notation ) and the like.

Optionally, after receiving the input information input by the user, the input information may be first subjected to preprocessing such as cleaning, word segmentation, part-of-speech tagging, and the like, and potential entities in the input information are identified, and the identified entities are standardized, for example, the spoken "domestic total production value" is mapped to a unified database identifier GDP, and then the standardized input information may be processed through the first large language model.

Optionally, after generating the plurality of subtasks, it may further output the same and normalized requirements to the interactive interface for user confirmation or adjustment.

Step S20, distributing a plurality of subtasks to other agents in a preset agent role pool through the product manager agent;

An agent refers to an artificial intelligence entity that can perceive an environment and take action to achieve a characteristic goal. The agent role pool refers to an agent cluster registered in advance, each agent has independent task processing capability, and is bound with a specific capability label for matching with user requirements during subtask allocation.

The intelligent agent role pool comprises a product manager intelligent agent, a data analysis and development intelligent agent and a visual designer intelligent agent, wherein the product manager intelligent agent is an intelligent agent with task scheduling and resource allocation algorithms, subtask allocation can be realized through capability label matching, the data analysis and development intelligent agent is an intelligent agent integrating various data analysis algorithms and data processing tools and is responsible for executing data related processing tasks such as data acquisition, cleaning, conversion, feature engineering, model selection, algorithm realization, data analysis and the like, and the visual designer intelligent agent is an intelligent agent with built-in visual design templates and graphic drawing tools to realize data visualization and is responsible for designing visual schemes including graph types, layouts, color matching and the like according to data and subtask requirements.

The first large language model converts the economic index visual demand information input by the user into a plurality of subtasks, and the subtasks are sent to the product manager agent, so that the product manager agent can match each subtask with the capability labels of the agents in the agent role pool, and any subtask is distributed to the agent with the highest matching score.

Alternatively, the plurality of subtasks may be assigned to other agents in the preset agent role pool through a product manager agent and a preset scheduler, wherein the scheduler is used to manage the above information of the product manager agent and the output information of the other agents.

For example, when a product manager agent is performing subtask allocation, long-term operation accumulation or system burst failure may cause partial data loss, and at this time, the scheduler may acquire the above information to recover lost task allocation data, so as to ensure continuity and accuracy of task allocation.

Optionally, the product manager agent determines the dependency relationship among the plurality of subtasks and distributes the plurality of subtasks to other agents in the agent role pool, and further, the other agents are controlled to execute the distributed subtasks according to the dependency relationship.

Step S30, determining a target data source through data analysis and development of the intelligent agent and the first subtasks allocated to the intelligent agent, and generating a data analysis result according to data in the target data source;

The first subtask is used to characterize the subtasks related to data processing assigned to data analysis and development agents, and generally includes explicit economic indicators, analysis requirements (such as growth rate), space-time constraints, and the like.

The target data source refers to a data access point determined according to the first subtask analysis, and may be a database, an API (Application Programming Interface ) provided by an official statistical structure, an unstructured document library, a real-time data stream, and the like, which is not particularly limited in this embodiment.

The data analysis result refers to information and conclusion obtained by searching, analyzing and/or predicting the target data source according to the first subtask through the data analysis and development agent. The data analysis and development agent can select an appropriate analysis model (for analyzing tasks such as relationship among economic indexes) or a prediction model (for tasks such as economic trend prediction and risk assessment) to analyze or predict the retrieved data according to the first subtask so as to generate a data analysis result.

The data analysis and development agent can search and match in a known data source list or database through analyzing the economic index, time range and other information contained in the first subtask to determine a target data source containing required data, further construct a data request instruction according to the first subtask and send the instruction to the target data source to obtain a data set meeting the requirement of the first subtask, further determine a data analysis method and model according to the data type and the requirement of the task, and process and analyze the retrieved data set by utilizing the determined data analysis method and model to generate a data analysis result.

In a possible embodiment, the step of determining the target data source in step S30 by data analysis and development of the agent and its assigned first subtask includes:

Step S31, determining a target economic entity according to the first subtask through data analysis and development agent;

The economic entity is an object of data association analysis, and may be economic indexes such as GDP and CPI, microscopic economic entities such as industry and enterprise, or economic events, economic policies, geographical areas, time dimensions, etc., which are not particularly limited in this embodiment.

The target economic entity can be obtained by performing semantic analysis on the first subtask through data analysis and development agents, and generally comprises index names, regions and time.

Step S32, determining target matching degree between a target economic entity and each data source in a preset data source library through data analysis and development of the intelligent agent;

the data source library refers to a knowledge base for structurally storing a plurality of pieces of data source information, wherein each entry records metadata such as names of data sources, API interfaces, economic index lists covered by the data sources (adopting standardized names), time granularity (year/season/month/day), geographic granularity (country/province/city), data update frequency, authority level of a data providing mechanism, data quality level and the like.

The target matching degree refers to the association strength between the target economic entity and each data source, and can be determined according to indexes such as semantic similarity, data dimension coverage, time granularity alignment degree and the like.

In one possible embodiment, step S32 includes:

Step S321, determining the standardized name of the target economic entity according to a preset economic knowledge graph through data analysis and development of the intelligent agent;

The economic knowledge graph refers to graph structure data formed by nodes (economic entities) and edges (relationships among the entities), node attributes comprise names, types, regions, service ranges and the like, and edge attributes comprise relationships of membership, cooperation, competition and the like. And the standardized names refer to standard names which are unique in the economic knowledge graph and are used for identifying a certain economic entity in a standardized manner.

Optionally, before step S321, the economic domain data may be obtained by using open APIs provided by a plurality of different economic data sources (including statistics annual, financial report, policy file, research paper, news report, industry analysis report, etc.), and then entity identification, relationship extraction and event extraction are performed on the economic domain data by using a large model, so as to obtain economic entities (including economic events) and economic relationships in the economic domain, and an economic knowledge graph is constructed based on the economic entities and economic relationships. The large model is obtained through economic text data fine adjustment marked with entities, relations and events.

Optionally, new data can be dynamically obtained from the economic data source, the knowledge extraction process is repeated, and the economic knowledge graph is dynamically updated and expanded according to the new knowledge obtained by extraction.

Optionally, in the dynamic construction process of the economic knowledge graph, part of the knowledge graph content can be extracted by a manual sampling inspection mode, and whether the problems of entity identification errors, inaccurate relation extraction and the like exist or not is checked, so that the accuracy of information in the economic knowledge graph is ensured.

Optionally, in the dynamic construction process of the economic knowledge graph, analysis processing can be performed according to economic field data extracted by a preset conflict detection and resolution rule. For example, when data extracted from different economic data sources conflict (e.g., different economic data sources have different records on the values of the same economic index), which knowledge is more reliable can be determined according to authority of the economic data sources (e.g., data of the authority statistics department is preferentially selected), freshness of the data (e.g., data with more recent selection time), and the economic knowledge graph can be updated accordingly.

It can be understood that by determining the standardized names of the target economic entities, the problem that the economic index names are inconsistent in expression under different sources and different scenes can be effectively solved, errors and repeated labor caused by name differences can be reduced, and the efficiency of the whole economic index analysis and visualization process can be improved.

Step S322, determining the keyword matching degree between the standardized name and the index item of each data source through data analysis and development of the intelligent agent;

the keyword Matching degree is the vocabulary overlapping degree of the index standardized name and the index item of the data source, and can be determined by a TF-IDF (Term Frequency-Inverse Document Frequency) algorithm, a BM25 (Best Matching 25) algorithm, and the like.

Step S323, determining the semantic matching degree between the input information and the index items of each data source through data analysis and development of the intelligent agent;

the semantic matching degree refers to the matching degree obtained by comparing the deep meaning of the concept expressed by the input information and the data source index item by using a natural language processing technology and a semantic analysis algorithm, and is used for measuring the consistency between the original index requirement and the data source index.

For any data source, the BERT (Bidirectional Encoder Representations from Transformers) model may be used to convert the input information and the index item of the data source into vectors, and calculate the cosine similarity between the two vectors as the semantic matching degree between the input information and the index item of the data source.

Step S324, determining target matching degree according to the keyword matching degree and the semantic matching degree through data analysis and development of the intelligent agent.

For any data source, the keyword matching degree and the semantic matching degree of the target economic entity and the data source can be weighted and summed according to a preset weight distribution rule, so that the target matching degree between the target economic entity and the data source is obtained.

It can be understood that the limitation of purely relying on name matching or semantic matching is overcome by a keyword matching and semantic matching mixed matching algorithm, the situation of data source misselection caused by similar names but inconsistent semantics or similar semantics but larger name difference is effectively avoided, the accuracy and reliability of data acquisition are improved, and the accuracy of economic index analysis is further improved.

And step S33, determining a target data source through data analysis and development of the intelligent agent and the matching degree of each target.

For example, the data analysis and development agent may select, from the database of data sources, the data source with the highest target match or meeting a certain match threshold or above as the target data source.

Optionally, in the case that there are multiple data sources that can provide the same economic index, the priority of each data source that can provide the same economic index may be determined according to preset priority rules such as authority score, data update timeliness, data granularity matching degree, API stability, and the like of the data sources, and the data source with the highest priority is determined as the target data source.

In this embodiment, the economic knowledge graph and the hybrid matching algorithm are combined to determine the target data source, so as to improve the accuracy of data source matching, and ensure that the economic index data meeting the user requirements can be obtained.

And S40, generating a visual output corresponding to the data analysis result through the visual designer agent and the second subtasks distributed to the visual designer agent.

The second subtask is used to characterize subtasks related to data visualization assigned by the visual designer agent, typically including requirements in terms of visualization form, style, elements, etc.

Visual output refers to data analysis results which are generated after the visual designer agent processes and are presented in an intuitive visual form, and the visual output comprises forms of interaction charts, analysis reports, large screen components and the like.

For example, the visual designer agent may determine the horizontal axis, the vertical axis and the corresponding visual form, such as a line graph, a pie chart, a stacked bar chart, etc., according to the second subtask, and assume that the visual form is determined to be the line graph, and further the visual designer agent draws the line graph with the year as the horizontal axis and the GDP value as the vertical axis by the data analysis result regarding the GDP change in the last five years and the built-in drawing tool and algorithm thereof, so as to reflect the change of the GDP value in each year. Meanwhile, visual designer intelligent agents can automatically optimize visual elements such as layout, color, fonts, labels and the like of charts according to potential audience and user preferences, and the professionals and the readability of visual output are ensured.

The visual designer agent and the visual designer agent are distributed to the second subtasks to generate a visual chart corresponding to the data analysis result, chart titles, legend descriptions and key insights are automatically generated according to chart contents, hard data stacking is avoided, further, the product manager agent can generate introduction and background introduction according to input information input by a user, meanwhile, the data analysis and development agent can generate key findings on economic indexes according to the data analysis method and the data analysis process, and the data analysis result is combined to generate conclusions and suggestions, and further, the product manager agent can combine the introduction, the background introduction, the data analysis method, the key findings, chart interpretation, conclusions and suggestions to generate an economic analysis report.

Optionally, before the visual output is generated by the visual designer agent, a target chart is selected from a preset chart library according to the data type of each data in the data analysis result and the visual preference setting of the user through a preset large model, and a preset number of candidate layout schemes with the target chart arranged are generated, and then, the target layout schemes are determined from the candidate layout schemes through the visual designer agent and the second subtask, and the data analysis result is filled into the chart of the target layout scheme, so that the visual output corresponding to the data analysis result is obtained.

Optionally, under the condition that the input information comprises a plurality of economic indexes, the visual designer agent can generate a corresponding visual chart aiming at each economic index, and further, the visual designer agent can arrange and layout the generated visual charts to generate a comprehensive interactive economic index instrument board, so that a user can conveniently acquire information in one stop.

An exemplary process diagram of the economic index visualization is provided in fig. 2, wherein the process diagram of the economic index visualization is provided in fig. 2, firstly, data input is provided for a visual designer agent, including data analysis results and user visualization preferences, then the visual designer agent performs visual design according to the data input, during which the visual designer agent can drive to provide various visual chart selections and layout schemes by means of a large model, then the visual designer agent determines final chart composition and target layout schemes from the various layout schemes provided by the large model, the chart composition can be a line chart, a GIS (Geographic Information System, a geographic information system) map, a multidimensional cylindrical chart, a compound chart and the like, the process diagram is not limited in particular, then data filling is performed according to the data analysis results, the visual chart is obtained, chart interpretation is generated according to chart content and data analysis results, and finally, the visual report and chart interpretation are combined according to a display form in a second sub-task, and final visual output is determined, wherein the display form can include a comprehensive panel, a data viewing board, a comprehensive large screen, an analysis form and the like.

In the embodiment, through cooperation between the large model and the multiple intelligent agents, a complete autonomous process from understanding of economic index requirements to final delivery is realized, and the large model and the intelligent agents can be automatically and seamlessly joined without manual intervention, so that the efficiency of economic index visualization is improved on the whole operation process.

In the second embodiment of the present application, the same or similar content as in the first embodiment of the present application may be referred to the description above, and will not be repeated. On this basis, the step of generating a data analysis result according to the data in the target data source in step S30 includes:

step S34, pulling target data from a target data source through data analysis and development of the intelligent agent;

Target data refers to a data set pulled from a target data source, and may include different data formats such as tables, JSON, and the like.

The data analysis and development agent can select the adaptive protocol through the type of the target data source, construct an API request and send the API request to the target data source, further receive the data stream returned by the target data source, store the data stream in the temporary buffer area and wait for the next processing.

In one possible embodiment, the agent role pool includes a data architect agent, after step S34, further comprising:

step S341, in the case that a plurality of target data sources exist, judging whether the index dimensions of the target data pulled from each target data source are consistent or not through a data architect agent;

The data architect agent is an agent with data structure analysis and dimension verification capability, generally integrates algorithms such as data blood edge analysis, data dimension alignment and the like, and is responsible for designing a data pipeline, determining a data storage scheme and planning data circulation.

The index dimension refers to a metadata attribute set of each field in the target data, including data types, units, and the like, for example, the "sales" field may be counted by "day/month" in different data sources.

Step S342, under the condition that the index dimensions of all the target data are inconsistent, determining the target dimensions through the data architect agent and the first subtask, and uniformly converting the index dimensions of all the target data into target dimensions;

the target dimension refers to a standard dimension that enables unified analysis and comparison of data.

The data architect agent analyzes the first subtask under the condition that the index dimensions of the target data are inconsistent, determines the target dimension expected to be displayed by the user from the index dimensions corresponding to the target data according to the user requirement, and further, converts the index dimensions of the target data into the target dimensions in a unified way by means of a conversion algorithm (such as reduction index, exchange rate conversion and the like) specialized in the economic field according to the dimension difference.

In step S343, each target data is integrated by the data architect agent.

For example, the data architect agent may integrate the target data according to the standard name of each field in the target data, e.g., integrate the data within three years and three years ago for the field according to the standard name for unified viewing and recall.

In the embodiment, the dimension conversion and the data integration are automatically performed by the data architect intelligent agent, the dimension difference and the data inconsistency among different data sources are eliminated, a unified and comprehensive data set is constructed, a reliable data base is provided for economic index analysis, and the accuracy of data analysis is improved.

Step S35, performing data cleaning on the target data through data analysis and development of the intelligent agent to obtain cleaned target data;

The data cleaning refers to the process of carrying out anomaly detection and correction on target data based on economic rules or economic knowledge patterns, and comprises data correction, outlier processing, missing value filling, data quality verification and the like.

The data analysis and development agent can perform semantic understanding and context association analysis on the target data through a large language model, identify and correct wrongly written words and inconsistent formats in the target data, automatically detect and process abnormal values through a statistical method and a machine learning algorithm (such as an isolated forest algorithm and the like), intelligently fill missing data based on time sequence prediction, a regression model or context association information, and perform logic consistency verification (such as GDP (graphic data protocol) cannot be smaller than CPI), data timeliness verification (data is guaranteed to be latest and available) and data integrity check through an economic knowledge graph.

It can be appreciated that the data cleaning can effectively remove noise, errors and repeated information in the data, and improve the accuracy, the integrity and the consistency of the target data so as to improve the accuracy of the economic index analysis.

And step S36, searching the cleaned target data according to the first subtask by the data analysis and development agent to generate a data analysis result.

The method comprises the steps of analyzing a first subtask by a data analysis and development agent, determining search conditions such as a time range and a field filtering rule, converting the search conditions into corresponding database query languages or search instructions, searching the cleaned target data to obtain search results, and calling a related analysis algorithm or model according to user requirements in the first subtask to analyze and process the search results to generate data analysis results.

In one possible embodiment, step S36 includes:

step S361, determining a target economic entity according to the first subtask through the data analysis and development agent;

The specific implementation of step S361 can refer to the specific implementation of step S31 in the above-mentioned first embodiment, and will not be described herein.

Step S362, determining the expansion information of the target economic entity according to the preset economic knowledge graph;

It should be noted that, step S362 may use the economic knowledge graph constructed in the first embodiment.

The extended information is obtained through economic knowledge graph reasoning and is related to other economic concepts, entities or information of the target economic entity.

The data analysis and development agent can automatically extend the target economic entity to synonyms, related entities or superior concepts thereof according to the entities and the relations in the economic knowledge graph, and obtain the extension information corresponding to the target economic entity so as to improve the retrieval recall rate. For example, the entity "digital economy" may be extended to "information industry", "big data", "artificial intelligence", etc.

Step S363, according to the expansion information, keyword retrieval is carried out on the cleaned target data to obtain a first retrieval result;

the first search result is used for representing the structured data obtained through keyword search.

The data development and analysis agent can perform keyword matching on different fields in the extension information and an index list of the cleaned target data through a preset text retrieval algorithm (such as a BM25 algorithm) to determine a matched target index, further extract data corresponding to the target index from the cleaned target data, and combine the target index and the corresponding data to obtain a first retrieval result.

It can be understood that the data related to the target economic entity can be further focused by searching the cleaned target data by utilizing the keywords in the expansion information, so as to improve the data analysis effect.

Step S364, converting the input information and the expansion information into text vectors by utilizing the trimmed large language model, wherein the trimmed large language model is obtained by trimming training data in the economic field;

the trimmed large language model refers to a specialized language model obtained by further training on training data in the economic field based on the original large language model. The text vector is a vector representation of semantic features extracted by the trimmed large language model for representing the input information and the expansion information.

Step S365, determining a second retrieval result according to the vector similarity between the text vector and each index item in the cleaned target data;

The vector similarity is used for measuring a quantization index of the similarity degree of two vectors in a high-dimensional space, and can be calculated by methods such as cosine similarity, euclidean distance and the like.

The second search result is used to characterize the structured data determined by the vector similarity calculation.

The data development and analysis agent can vectorize input information and expansion information by utilizing a trimmed large language model to obtain text vectors, vectorize all index items in the cleaned target data by utilizing the trimmed large language model to obtain a plurality of index vectors, calculate vector similarity between the text vectors and all the index vectors through a cosine similarity algorithm, determine index items corresponding to index vectors with vector similarity exceeding a preset threshold as target indexes, extract data corresponding to the target indexes from the cleaned target data, and combine the target indexes and the corresponding data to obtain a second retrieval result.

It can be understood that the second search result is determined by calculating the vector similarity between the text vector and the target data index item, so that accurate data search based on semantic understanding is realized, and the problem that semantic related but literal unmatched data may be missed in traditional keyword search is solved.

Step S366, determining a target search result according to the first search result and the second search result, and generating a data analysis result according to the target search result.

The data development and analysis agent can combine the first search result and the second search result according to the index name to remove repeated index items and obtain a target search result, and then call a data analysis algorithm to analyze the target search result to generate a data analysis result.

Optionally, after the first search result and the second search result are obtained, each index item may be ranked according to the matching degree of each index item and the input information, authority of the corresponding data source, release time, connectivity in the economic knowledge graph (i.e. association strength between the economic knowledge graph and other important economic concepts), and the like, and each index item may be combined according to the ranking to obtain the target search result. And in the subsequent visualization process, the index items which meet the requirements of the user and have strong relevance can be displayed preferentially, so that the visualization experience of the user is improved.

An exemplary process diagram of economic data processing is provided in fig. 3, wherein the process diagram of economic data processing is that firstly, data input is determined, data can be read from a data source library, each data source can be data from a national statistical bureau, an intra-provincial data interface, an industry report, a real-time financial data stream and the like, then data source discovery and connection are performed through data analysis and development agents, after the target data source is determined, data are pulled from the target data source, and under the condition that a plurality of target data sources exist, dimension conversion and integration are performed on target data of different data sources, a complete target data (which can be presented in the form of a data set/a data table and the like) is obtained through integration, then data cleaning and verification are performed on the integrated target data, including data correction, abnormal value processing, missing value filling, data quality verification and the like, and further data retrieval is performed by combining knowledge graphs, wherein the knowledge graphs can be obtained through integration of a plurality of knowledge graphs, such as economic knowledge base knowledge, industry private domain knowledge and the like, the data retrieval can comprise keyword retrieval and the retrieval vector retrieval and the retrieval result of the economic index is the retrieval result of the economic index obtained by combining the two data with the output of the economic index.

In the embodiment, by combining the mixed search strategy of keyword search and vector search, the data universality is ensured, the data correlation and accuracy are ensured, the comprehensiveness and the accuracy of search results are improved, and the accuracy of economic index analysis is improved.

In one possible embodiment, the agent role pool includes a quality assurance agent, which further includes, after step S30:

Step A10, judging whether the predicted deviation between the predicted value in the data analysis result and the obtained actual value exceeds a preset threshold value or not through a quality assurance agent;

The quality assurance agent is an agent with the capabilities of state sensing, deviation calculation, root cause analysis, model optimization and the like, and is responsible for comprehensive testing and verification of data accuracy, model validity, visual presentation effect, system functionality and the like.

Optionally, in the whole life cycle of the visual analysis of the economic index, the whole visual arrangement flow can be monitored in real time through the quality assurance agent cooperation, including monitoring the connection state of a data source, ensuring the stability of a data interface and no interruption of data flow, monitoring the efficiency and resource consumption of data cleaning, conversion and integration processes so as to realize the monitoring of data processing performance, tracking the running speed and accuracy of a prediction model or an analysis algorithm, monitoring the visual generation efficiency, ensuring the generation speed of a chart and a report to meet the requirements, monitoring the use condition of resources such as a CPU (Central Processing Unit, a central processing unit), a memory, a storage and the like.

Illustratively, the quality assurance agent may evaluate the quality of the data by monitoring the integrity, accuracy, consistency, timeliness, etc. of the target data.

The quality assurance agent may evaluate a prediction model called by the data development and analysis agent in the data analysis process through indexes such as R-squared, MAE (Mean Absolute Error ), RMSE (Root Mean Squared Error, root mean square error), MAPE (Mean Absolute Percentage Error ), and the like, and evaluate an analysis model called by the data development and analysis agent in the data analysis process through indexes such as model fitting goodness, significance level, residual error, and the like. And then the back propagation algorithm and the judgment result can be utilized to adjust the prediction model or the analysis model or replace the model called by the data development and analysis agent.

And step A20, under the condition that the predicted deviation exceeds a preset threshold value, adjusting the intelligent agent parameters of the data analysis and development intelligent agent through a preset reinforcement learning strategy.

The prediction bias refers to an absolute or relative error between the model predicted value and the true value.

The reinforcement learning strategy is a machine learning method for updating an agent behavior strategy (agent parameters) according to a feedback reward signal through interaction of an agent and an environment, and comprises a PPO (Proximal Policy Optimization, near-end strategy optimization) algorithm, a SAC (Soft Actor-Critic) algorithm and the like.

The intelligent agent parameter definition data analyzes and develops intelligent agent behaviors and internal variables of decision logic, and can comprise called models, model parameters, feature weights, time attenuation factors and the like.

The method comprises the steps of determining a prediction deviation by comparing a prediction value in a data analysis result with a currently acquired time at regular time by a quality assurance agent under the condition that the data analysis and development agent invokes a prediction model to generate the data analysis result, generating a punishment signal according to the prediction deviation under the condition that the prediction deviation exceeds a preset threshold value, and adjusting the data analysis and development agent parameters, such as model parameters or invoked prediction models, through a PPO algorithm and the punishment signal.

In the embodiment, the parameters of the intelligent agent are automatically adjusted through the reinforcement learning strategy, so that the data analysis and development intelligent agent automatically optimizes the self behavior when the prediction deviation is overlarge, the accuracy and the reliability of prediction are gradually improved, the manual intervention is reduced, the efficiency of economic index analysis is ensured, and the reliability of the economic index analysis is improved.

In one possible embodiment, the method further comprises:

and step A30, uniformly storing the input information and the output information of all the agents in the agent role pool so as to ensure that the contexts are shared among all the agents in the agent role pool.

For example, any agent in the agent role pool can be composed of independent large language model examples, when the corresponding assigned subtasks are executed, independent analysis can be carried out in the large language models, and when the subtasks are executed, corresponding output information is stored in a preset shared database so that other agents can carry out data retrieval.

Optionally, for any agent in the agent role pool, when a conflict (such as lack of file creation authority) is detected, timing is started, and communication is performed with other agents in the agent role pool to solve the conflict, such as inquiring a product manager whether the agent can not create a new file, and when the conflict is solved, timing is stopped, but if the time when the conflict is detected exceeds a preset time threshold, alarm information can be output to seek help from a user.

An exemplary process diagram of multi-agent collaboration and communication is provided in fig. 4, in which, firstly, input information is processed through a first large language model to determine standardization requirements and a plurality of subtasks, the input information, the standardization requirements and the subtasks are all stored in a unified shared memory, then, subtasks are distributed to different agents through a product manager agent, corresponding subtask distribution information is also stored in the shared memory, then, each agent executes the distributed subtasks, for example, a data analysis and development agent can determine a target data source according to the distributed subtasks and draw data for data retrieval and analysis, after the data of the different data sources are drawn, the data analysis and development agent and the output information and communication context corresponding to the data architect agent are all stored in the shared memory for other agents to call, and then, a visual design agent can read the data from the sub-tasks and the shared memory according to the design agent and the visual design and output the visual design result to the visual analysis and the final analysis result. In the whole life cycle of visual output, the execution process and data of other intelligent agents can be monitored and comprehensively evaluated by the quality assurance intelligent agents, and intelligent agent parameters of all intelligent agents are reversely adjusted according to evaluation results.

In the embodiment, the information circulation among the intelligent agents is realized by uniformly storing the input information and the output information of the intelligent agents, so that the problem of repeated labor and inconsistent information is avoided, the continuity of the whole analysis flow is ensured, and the efficiency of economic index visualization and the reliability of analysis results are improved.

In the third embodiment of the present application, the same or similar contents as those of the first and second embodiments of the present application can be referred to the description above, and the description thereof will not be repeated. On this basis, the step of generating normalized requirements in step S10 by processing the received input information through the first large language model includes:

S11, processing input information through a first large language model, and determining user intention and semantic slots corresponding to the user intention;

The user intention refers to a user core request extracted through a large language model, the semantic slot refers to a key parameter to be filled under the user intention, and the key parameter is autonomously determined by the large language model according to training data.

In one possible embodiment, before step S10, further includes:

Step S01, intention recognition is carried out on preset training data through a second large language model, and a soft target of the training data is obtained;

Soft Target refers to the probability distribution of the teacher model output during knowledge distillation (Knowledge Distillation), including the probability distribution of the corresponding intent, entity type, and relationship type for each word or phrase in the training data, rather than a single category label.

The second large language model characterizes a teacher model for outputting a soft target, which may be an original large language model, such as DeepSeek (deep thinking), qwen (thousand questions), or the like, or a large language model obtained after fine tuning training data in the economic field, which is not particularly limited in this embodiment.

And step S02, training the third large language model through the training data and the soft target to obtain a first large language model.

The third large language model characterizes the learning model in the knowledge distillation process, which is typically set as the original large language model.

The training data and the soft targets are input into a training framework to train a third large language model so as to learn how to generate intention probability distribution similar to the soft targets according to input texts, namely learning experience of a teacher model (the second large language model) in generalization and fuzzy information processing, and after multiple rounds of iterative optimization, the performance of the third large language model reaches the expected or loss function convergence to obtain a first large language model for processing actual user input.

For example, the input information may be processed through the first large language model after distillation training, and for the input information "i want to see the trend of the GDP growth rate and CPI index in nearly five years in china, it is preferable to be able to compare and display in a graph", which corresponds to the identified user intention as "economic index trend comparison", and the semantic slots under the user intention include economic index, region, time range, visual form, and the like.

In this embodiment, the first large language model is obtained through training in a knowledge distillation manner, so that not only can economic entities be determined from the input information, but also the relationship between different entities can be identified, thereby more accurately identifying the user intention corresponding to the input information.

And step S12, extracting data content from the input information to fill semantic slots, so as to obtain normalized requirements.

The data content refers to the entity or attribute value associated with the semantic slot in the input information.

Illustratively, the data content can be directly filled in after being extracted, and the data content can be filled in after being inferred through the first large language model. For example, for the input information "I want to see the trend of GDP growth rate and CPI index of nearly five years in China", the data content corresponding to the regional slot is determined to be "China" and can be directly filled in, while the data content corresponding to the time range slot is determined to be "nearly five years", and specific year information (such as 2020-2025) can be determined through a large language model and then filled in. And combining the user intention and the filled semantic slots to obtain the normalized requirements corresponding to the input information.

In this embodiment, a first large language model is obtained through training of a knowledge distillation technology, and intention recognition is performed through the first large language model to generate normalized requirements, so that accuracy of user intention recognition is improved, the whole economic index visualization process is ensured to be executed based on accurate user requirements, user intervention is reduced, and efficiency of economic index visualization and accuracy of economic index analysis results are improved.

For an exemplary understanding of the implementation flow of the large-model multi-agent collaborative economic index visual arrangement method according to the first embodiment, please refer to fig. 5, fig. 5 provides a general flow chart of multi-agent collaborative, firstly, an interactive interface is provided, a user can input the economic analysis requirements and visual targets through a natural language text box or a voice input interface in the interactive interface, then, the user input is analyzed through an intelligent intention and requirement analysis module, a preliminary normalized requirement and task decomposition plan are output to the interactive interface for user confirmation or fine adjustment, further, the normalized requirement and a plurality of sub-tasks determined by the user are sent to the multi-agent collaborative management and arrangement module, task allocation is performed through the product manager, data source mining and data retrieval and data analysis are performed through the data analysis and development of the multi-agent collaborative intelligent index, the user can perform dimension conversion and integration on data of different data sources through the data architect, visual chart selection and visual layout are performed through the visual design intelligent agent, and final layout are output, the visual normalized requirement and task decomposition plan is provided to the interactive interface for user confirmation or fine adjustment, furthermore, the normalized requirement and a plurality of sub-tasks are sent to the multi-agent collaborative management and arrangement module through the product manager intelligent agent, the data analysis and development intelligent agent is used for mining target data source through the data retrieval and data analysis, the data architect data conversion and data analysis are performed through the data architect, the dimensional conversion and integration on data of different data sources, the data of the data source and the data of the data may be processed through the data architect and the data, and the data architect and the intelligent analyzer intelligent agent is processed through the economic index graph and the economic index analysis module, and the economic index analysis module and the economic index management and economic index management, so that the user knows the current project progress. Meanwhile, the multi-agent collaborative management and arrangement module also provides quality assurance agents to continuously monitor the performance of the visual output process and the stability of the underlying data source, and optimize each agent according to the improvement suggestion of the issued result by the user. In addition, in the multi-agent collaborative management and orchestration module, when an agent detects a conflict that is not resolved for a long period of time or an agreement is not reached inside the agent, a request may be issued to a human supervisor (user) and the conflict or negotiation may be resolved according to the user suggestion. Finally, after the complete visual output is displayed on the interactive interface, a user can export the visual result into the forms of documents, pictures and the like through one key by triggering operation on the release control, or deploy the visual result into an internal data platform for other people to review and interact.

It should be noted that the foregoing examples are only for understanding the present application, and do not constitute any limitation on the method for visualizing and arranging economic indicators based on large-model multi-agent collaboration, and it is within the scope of the present application to perform more forms of simple transformations based on the technical concept.

The large-model multi-intelligent collaborative economic index visual arrangement method is not only limited in the economic index visual field, but also can be widely applied to various scenes needing high-efficiency and autonomous management of complex information processing and decision processes, for example, in a supply chain and logistics optimization system, an agent team can cooperatively analyze data (such as production, transportation, inventory and sales) from different links, autonomously identify bottlenecks and generate an optimization scheme and a visual report, in smart city construction, the agent team can autonomously analyze and visualize city operation data (such as traffic flow, environmental index, public safety event and population density), assist a city manager in understanding city operation conditions, predict development trend, assist in formulating city planning and emergency response strategies, and can also be applied to scenes such as scientific project management and data analysis, financial risk management and investment analysis, intelligent marketing and customer relationship management, and the like, so that management efficiency, accuracy and intelligent level in each scene are improved.

The embodiment of the application also provides an economic index visual arrangement system based on large-model multi-agent cooperation, referring to fig. 6, the economic index visual arrangement system based on large-model multi-agent cooperation comprises:

The requirement analysis module 10 is configured to process the received input information through the first large language model, generate a normalized requirement, and decompose the normalized requirement into a plurality of subtasks;

the task allocation module 20 is configured to allocate, through a product manager agent, a plurality of subtasks to other agents in a preset agent role pool, where the agent role pool includes data analysis and development agents and visual designer agents;

the data analysis module 30 is configured to determine a target data source by data analysis and development of the agent and the first subtasks allocated thereto, and generate a data analysis result according to data in the target data source;

and the data visualization module 40 is used for generating a visual output corresponding to the data analysis result through the visual designer agent and the second subtasks allocated to the designer agent.

According to the large-model multi-agent cooperation-based economic index visualization arrangement system provided by the embodiment of the application, the technical problem of how to improve the efficiency of economic index visualization can be solved by adopting the large-model multi-agent cooperation-based economic index visualization arrangement method in the embodiment. Compared with the prior art, the economic index visual arrangement system based on large-model multi-agent cooperation has the same beneficial effects as the economic index visual arrangement method based on large-model multi-agent cooperation provided by the embodiment, and other technical features in the economic index visual arrangement system based on large-model multi-agent cooperation are the same as the features disclosed by the method of the embodiment, and are not repeated herein.

The embodiment of the application provides electronic equipment, which comprises at least one processor and a memory in communication connection with the at least one processor, wherein the memory stores instructions which 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 index visualization arrangement method based on large-model multi-agent cooperation in the first embodiment.

Referring now to fig. 7, a schematic diagram of an electronic device suitable for use in implementing embodiments of the present application is shown. The electronic device in the embodiment of the present application may include, but is not limited to, a mobile terminal such as a mobile phone, a notebook computer, a digital broadcast receiver, a PDA (Personal DIGITAL ASSISTANT: personal digital assistant), a PAD (Portable Application Description: tablet computer), a PMP (Portable MEDIA PLAYER: portable multimedia player), an in-vehicle terminal (e.g., an in-vehicle navigation terminal), and the like, a fixed terminal such as a digital TV, a desktop computer, and the like. The electronic device shown in fig. 7 is only an example and should not be construed as limiting the functionality and scope of use of the embodiments of the application.

As shown in fig. 7, the electronic device may include a processing means 1001 (e.g., a central processing unit, a graphics processor, etc.) which may perform various appropriate actions and processes according to a program stored in a read only memory 1002 or a program loaded from a storage means 1003 into a random access memory 1004. In the random access memory 1004, various programs and data necessary for the operation of the electronic device are also stored. The processing device 1001, the read only memory 1002, and the random access memory 1004 are connected to each other by a bus 1005. An input/output interface 1006 is also connected to the bus. In general, a system including an input device 1007 such as a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc., an output device 1008 including a Liquid crystal display (LCD: liquid CRYSTAL DISPLAY), a speaker, a vibrator, etc., a storage device 1003 including a magnetic tape, a hard disk, etc., and a communication device 1009 may be connected to the input/output interface 1006. The communication means 1009 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. While electronic devices having various systems are shown in the figures, it should be understood that not all of the illustrated systems are required to be implemented or provided. More or fewer systems may alternatively be implemented or provided.

In particular, according to embodiments of the present disclosure, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method shown in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via a communication device, or installed from the storage device 1003, or installed from the read only memory 1002. The above-described functions defined in the method of the disclosed embodiment of the application are performed when the computer program is executed by the processing device 1001.

The electronic equipment provided by the embodiment of the application adopts the economic index visualization arrangement method based on large-model multi-agent cooperation in the embodiment, so that the technical problem of how to improve the efficiency of economic index visualization can be solved. Compared with the prior art, the electronic equipment provided by the application has the same beneficial effects as the economic index visual arrangement method based on large-model multi-agent cooperation provided by the embodiment, and other technical characteristics in the electronic equipment are the same as the characteristics disclosed by the method of the previous embodiment, and are not repeated herein.

It is to be understood that portions of the present disclosure may be implemented in hardware, software, firmware, or a combination thereof. In the description of the above embodiments, particular features, structures, materials, or characteristics may be combined in any suitable manner in any one or more embodiments or examples.

The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Embodiments of the present application provide a computer readable storage medium having computer readable program instructions (i.e., a computer program) stored thereon, where the computer readable program instructions are configured to perform the economic indicator visualization arrangement method based on large model multi-agent collaboration in the above embodiments.

The computer readable storage medium provided by the embodiments of the present application may be, for example, a usb disk, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system or device, or a combination of any of the foregoing. More specific examples of a computer-readable storage medium may include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access Memory (RAM: random Access Memory), a Read-Only Memory (ROM), an erasable programmable Read-Only Memory (EPROM: erasable Programmable Read Only Memory or flash Memory), an optical fiber, a portable compact disc Read-Only Memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this embodiment, the computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system or device. Program code embodied on a computer readable storage medium may be transmitted using any appropriate medium, including but not limited to electrical wiring, fiber optic cable, RF (Radio Frequency) and the like, or any suitable combination of the foregoing.

The computer readable storage medium may be included in the electronic device or may exist alone without being incorporated into the electronic device.

The computer-readable storage medium carries one or more programs which, when executed by an electronic device, cause the electronic device to perform the functions described above as defined in the methods of the disclosed embodiments of the application.

Computer program code for carrying out operations of the present application may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of remote computers, the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN: local Area Network) or a wide area network (WAN: wide Area Network), or may be connected to an external computer (for example, through the Internet using an Internet service provider).

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The modules involved in the embodiments of the present application may be implemented in software or in hardware. Wherein the name of the module does not constitute a limitation of the unit itself in some cases.

The readable storage medium provided by the embodiment of the application is a computer readable storage medium, and the computer readable storage medium stores computer readable program instructions (namely computer program) for executing the economic index visualization arrangement method based on large-model multi-agent cooperation, so that the technical problem of how to improve the efficiency of the economic index visualization can be solved. Compared with the prior art, the beneficial effects of the computer readable storage medium provided by the application are the same as those of the economic index visual arrangement method based on large-model multi-agent cooperation provided by the embodiment, and are not repeated here.

The embodiment of the application also provides a computer program product, which comprises a computer program, wherein the computer program realizes the steps of the economic index visualization arrangement method based on large-model multi-agent cooperation when being executed by a processor.

The computer program product provided by the embodiment of the application can solve the technical problem of how to improve the efficiency of economic index visualization. Compared with the prior art, the beneficial effects of the computer program product provided by the application are the same as those of the economic index visualization arrangement method based on large-model multi-agent cooperation provided by the embodiment, and are not repeated here.

The foregoing description is only a partial embodiment of the present application, and is not intended to limit the scope of the present application, and all the equivalent structural changes made by the description and the accompanying drawings under the technical concept of the present application, or the direct/indirect application in other related technical fields are included in the scope of the present application.

Claims (8)

1. The large-model multi-agent cooperation-based economic index visual arrangement method is characterized by comprising the following steps of:

processing received input information through a first large language model, generating a normalized demand, and decomposing the normalized demand into a plurality of subtasks, wherein the input information is visual demand information aiming at economic indexes and input by a user;

Distributing the subtasks to other agents in a preset agent role pool through a product manager agent, wherein the agent role pool comprises data analysis and development agents and visual designer agents, the product manager agent is an agent with task scheduling and resource distribution algorithms, subtask distribution is realized through capability tag matching, and the data analysis and development agents are agents integrating various data analysis algorithms and data processing tools and are responsible for executing data-related processing tasks;

determining a target economic entity through data analysis and development of the intelligent agent according to the first subtask distributed to the intelligent agent;

Determining the standardized name of the target economic entity according to a preset economic knowledge graph through the data analysis and development intelligent agent;

Determining the keyword matching degree between the standardized name and index items of all data sources in a preset data source library through the data analysis and development agent;

determining semantic matching degree between the input information and index items of each data source through the data analysis and development agent;

determining target matching degree between the target economic entity and each data source according to the keyword matching degree and the semantic matching degree through the data analysis and development agent;

determining a target data source through the data analysis and development agent and the matching degree of each target;

Generating a data analysis result according to the data in the target data source;

And generating visual output corresponding to the data analysis result through the visual designer agent and the second subtasks distributed to the visual designer agent, wherein the visual designer agent is an agent with built-in visual design templates and graphic drawing tools for realizing data visualization and is responsible for designing a visual scheme according to data and subtask requirements.

2. The method for visualizing orchestration of economic indicators based on large model multi-agent collaboration as in claim 1, wherein said step of generating data analysis results from data in said target data source comprises:

Pulling target data from the target data source through the data analysis and development agent;

performing data cleaning on the target data through the data analysis and development intelligent agent to obtain cleaned target data;

And searching the cleaned target data according to the first subtask by the data analysis and development agent to generate a data analysis result.

3. The method for visualizing orchestration of economic indicators based on large-model multi-agent collaboration according to claim 2, wherein said step of retrieving said cleaned target data by said data analysis and development agent according to said first subtask, generating a data analysis result comprises:

determining a target economic entity according to the first subtask through the data analysis and development agent;

according to a preset economic knowledge graph, determining the expansion information of the target economic entity;

according to the expansion information, keyword retrieval is carried out on the cleaned target data, and a first retrieval result is obtained;

Converting the input information and the expansion information into text vectors by utilizing a trimmed large language model, wherein the trimmed large language model is obtained by trimming training data in the economic field;

determining a second retrieval result according to the vector similarity between the text vector and each index item in the cleaned target data;

And determining a target search result according to the first search result and the second search result, and generating a data analysis result according to the target search result.

4. The large model multi-agent collaborative based economic indicators visualization orchestration method according to claim 2, wherein the agent role pool comprises data architect agents, after the step of pulling target data from the target data source, further comprising:

In the case that a plurality of target data sources exist, judging whether index dimensions of target data pulled from each target data source are consistent or not through the data architect agent;

under the condition that the index dimensions of the target data are inconsistent, determining target dimensions through the data architect agent and the first subtask, and uniformly converting the index dimensions of the target data into the target dimensions;

integrating each target data by the data architect agent.

5. The large model multi-agent collaborative based economic index visualization orchestration method according to claim 1, wherein the agent role pool comprises quality assurance agents, after the step of generating data analysis results, further comprising:

Judging whether the predicted deviation between the predicted value in the data analysis result and the obtained actual value exceeds a preset threshold value or not through the quality assurance agent;

and under the condition that the prediction deviation exceeds the preset threshold value, the data analysis and the intelligent agent parameter of the intelligent agent development are adjusted through a preset reinforcement learning strategy.

6. The large model multi-agent collaboration-based economic indicator visualization orchestration method according to claim 1, wherein the method further comprises:

And uniformly storing the input information and the output information of all the agents in the agent role pool so as to ensure that the context is shared among all the agents in the agent role pool.

7. The method for visualizing orchestration of economic indicators based on large-model multi-agent collaboration according to claim 1, wherein said step of processing received input information through a first large language model to generate normalized requirements comprises:

processing the input information through the first large language model, and determining user intention and semantic slots corresponding to the user intention;

And filling the semantic slots according to the input information to obtain the normalized requirements.

8. The method for visualizing orchestration of economic indicators based on large model multi-agent collaboration of claim 7, further comprising, prior to said step of generating normalized requirements by processing received input information through a first large language model:

Performing intention recognition on preset training data through a second large language model to obtain a soft target of the training data;

And training a third large language model through the training data and the soft target to obtain the first large language model.