Enterprise diagnosis and operation report joint generation method and system based on multi-agent cooperation

Abstract

The application relates to a kind of enterprise diagnosis and operation report joint generation method and system based on multi-agent cooperation, comprising the following steps: obtaining the multidimensional operation data of target enterprise and carrying out time sequence unified processing, generating analysis dataset;Parallelly call several corresponding professional analysis agents;Synchronize the analysis dataset into professional analysis agent, obtain diagnostic score result;The obtained diagnostic score result is fused and conflict is resolved, and comprehensive diagnostic conclusion is generated;According to the comprehensive diagnostic conclusion, match text narrative template, data visualization chart component and conclusion suggestion corpus, and fill and associate through report integration agent, generate enterprise operation diagnosis report;In summary, by obtaining multidimensional operation data, carrying out time sequence unified processing and calling professional analysis agent parallel analysis, fusing diagnostic result to generate comprehensive conclusion and matching template to generate report, the problem of diagnosis and report generation in the prior art can be solved.

Description

Technical Field

[0001] This application relates to the technical fields of artificial intelligence and enterprise management informatization, and in particular to a method and system for generating enterprise diagnostic and operational reports based on multi-agent collaboration. Background Technology

[0002] In existing technologies, enterprise operation diagnosis and report generation typically rely on two main methods. One is manual analysis by expert teams, which heavily depends on personal experience, is time-consuming, and struggles to ensure consistency in evaluation standards across different-sized enterprises and business dimensions, often resulting in insufficient objectivity and comprehensiveness in diagnostic conclusions. The other method utilizes a single automated analysis model or system, which, while efficient, typically has fixed and limited analytical dimensions, making it difficult to conduct simultaneous and in-depth cross-analysis of multiple professional areas such as enterprise finance, market, supply chain, and organizational effectiveness. Furthermore, existing technical solutions generally suffer from a disconnect between the diagnosis and report generation stages. Analysis results are often presented as discrete data reports or simple conclusions, while generating a comprehensive diagnostic report with a complete structure and targeted recommendations still requires extensive manual work in data processing, chart creation, and text compilation. This not only lengthens the overall cycle from diagnosis to final report output but also makes the report quality susceptible to the subjective factors of the report writer, hindering a rapid, accurate, and standardized response to the dynamic diagnostic needs of enterprises. Summary of the Invention

[0003] To address the aforementioned shortcomings, this application provides a method and system for generating enterprise diagnostic and operational reports based on multi-agent collaboration.

[0004] The above-mentioned objective of this application is achieved through the following technical solution:

[0005] A method for generating enterprise diagnostic and operational reports based on multi-agent collaboration includes the following steps:

[0006] In response to enterprise diagnostic requests, obtain multi-dimensional operational data of the target enterprise corresponding to the enterprise diagnostic request;

[0007] Perform time-series unified processing on multidimensional operational data to generate an analysis dataset, and invoke several corresponding professional analytical agents in parallel based on enterprise diagnostic requests;

[0008] The analysis dataset is synchronously input into a professional analytical agent to obtain several diagnostic scores, including quantitative indicators, problem localization, and qualitative evaluation.

[0009] The obtained diagnostic scores are fused and conflict-resolved according to the preset integration logic to generate a comprehensive diagnostic conclusion, which includes core problem attribution, overall health score and priority improvement item list.

[0010] Based on the comprehensive diagnostic conclusions, the corresponding text description templates, data visualization chart components, and conclusion suggestion corpora are matched from the pre-set report knowledge base;

[0011] By integrating the report with an intelligent agent, the comprehensive diagnostic conclusions are populated and correlated with the selected text narrative templates, data visualization chart components, and conclusion suggestion corpora to generate an enterprise operation diagnostic report.

[0012] The second objective of this invention is achieved through the following technical solution:

[0013] A multi-agent collaborative enterprise diagnostic and operational report generation system includes:

[0014] The data acquisition module is used to respond to enterprise diagnostic requests and acquire multi-dimensional operational data of the target enterprise corresponding to the enterprise diagnostic request.

[0015] The intelligent agent invocation module is used to perform time-series unified processing of multi-dimensional operational data, generate analysis datasets, and invoke several corresponding professional analytical intelligent agents in parallel based on enterprise diagnostic requests.

[0016] The sub-result generation module is used to synchronously input the analysis dataset into the professional analysis agent to obtain several diagnostic sub-results, including quantitative indicators, problem localization, and qualitative evaluation.

[0017] The diagnostic conclusion generation module is used to integrate and resolve conflicts in the obtained diagnostic scores according to a preset integration logic to generate a comprehensive diagnostic conclusion, which includes core problem attribution, overall health score and priority improvement item list.

[0018] The template component matching module is used to match the corresponding text description template, data visualization chart component, and conclusion suggestion corpus from the pre-set report knowledge base based on the comprehensive diagnostic conclusion;

[0019] The report generation module is used to fill in and associate the comprehensive diagnostic conclusions with the selected text description templates, data visualization chart components, and conclusion suggestion corpora through the report integration intelligent agent, and generate an enterprise operation diagnostic report.

[0020] This application also relates to a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method for generating enterprise diagnostic and operational reports based on multi-agent collaboration.

[0021] This application also relates to a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for generating enterprise diagnostic and operational reports based on multi-agent collaboration.

[0022] In summary, the enterprise diagnosis and operation report generation method and system based on multi-agent collaboration provided in this application obtains multi-dimensional operation data by responding to enterprise diagnosis requests, performs unified time-series processing, calls professional analytical agents for parallel analysis, integrates diagnosis results to generate comprehensive conclusions, and automatically matches templates to generate reports. This solves the problem of the separation between diagnosis and report generation in existing technologies, and achieves efficient and objective diagnosis and report generation. It has the ability to perform in-depth analysis of multi-dimensional data and automatically generate comprehensive operation reports, thereby improving diagnosis efficiency and report quality and reducing manual intervention. Attached Figure Description

[0023] Figure 1 This is a flowchart of one step of an embodiment of the enterprise diagnosis and operation report generation method based on multi-agent collaboration in this application;

[0024] Figure 2 This is a flowchart illustrating the implementation of the multi-agent collaborative enterprise diagnosis and operation report generation method of this application;

[0025] Figure 3 This is a flowchart of step S10 in the embodiment of the enterprise diagnosis and operation report linkage generation method based on multi-agent collaboration in this application;

[0026] Figure 4 This is a flowchart of step S20 in the embodiment of the enterprise diagnosis and operation report generation method based on multi-agent collaboration in this application. Detailed Implementation

[0027] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0028] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0029] Existing enterprise operation diagnosis and report generation solutions suffer from a technical disconnect between the diagnosis and report generation stages. Specifically, the analytical results generated during the diagnosis process are typically presented as discrete data reports or simple conclusions, lacking an automatic mechanism to connect with the report generation stage. Furthermore, existing analytical dimensions are often limited by single automated models or human experience, making it difficult to achieve simultaneous cross-disciplinary assessments across multiple professional fields such as finance, marketing, supply chain, and organizational effectiveness. This results in deficiencies in the objectivity and comprehensiveness of diagnostic conclusions. In particular, because the quantitative indicators and problem identification output from the diagnosis stage are not structurally integrated, the qualitative evaluation lacks multi-dimensional evidence support, failing to automatically generate comprehensive conclusions that include core issues and key improvement items. This impacts diagnostic efficiency, report generation speed, and the consistency of evaluation standards, making it difficult to adapt to the dynamically changing diagnostic needs of enterprises.

[0030] For example, in a quarterly operational diagnostic scenario for a manufacturing company, it is necessary to process financial transaction data, production record data, and upstream and downstream supply chain data simultaneously. Under current technology, the financial analysis module will independently output cost indicator reports, and the supply chain analysis module will generate supplier delay risk assessments, but the results of the two are difficult to correlate effectively. When financial data shows an abnormal increase in raw material costs and supply chain data shows an extension of the delivery cycle of key suppliers, the existing technology makes it difficult to identify the causal relationship, which may lead to contradictory diagnostic results in problem localization. Furthermore, in the report generation stage, the existing system usually requires manual organization of discrete data into structured reports. In this process, because the quantitative indicators output by each module are not uniformly mapped to a standard measurement space, and the qualitative comments are not aligned at the semantic level, there are logical gaps in the report content, affecting the timeliness and accuracy of decision support.

[0031] In one embodiment, this application discloses a method for generating enterprise diagnostic and operational reports based on multi-agent collaboration, such as... Figures 1-2 As shown, the specific steps include the following:

[0032] S10: In response to an enterprise diagnostic request, obtain multi-dimensional operational data of the target enterprise corresponding to the enterprise diagnostic request;

[0033] In this embodiment, multi-agent collaboration refers to multiple software entities or systems with specific functions and decision-making capabilities working together to complete complex tasks through mutual communication, coordination, and cooperation. In this embodiment, different agents are responsible for different professional areas of enterprise diagnosis, and the comprehensiveness and depth of the diagnosis are improved through mutual collaboration. An enterprise diagnosis request is an instruction to analyze and evaluate the operational status of a specific enterprise. The enterprise diagnosis request typically includes parameters such as the target enterprise identifier, the scope of diagnosis, the business areas of interest, and the time period. Multidimensional operational data refers to a data set collected from multiple data sources inside and outside the enterprise, covering all aspects of enterprise operation. Multidimensional operational data includes financial data, production data, market data, human resource data, supply chain data, etc.

[0034] S20: Perform time-series unified processing on multi-dimensional operational data to generate an analysis dataset, and invoke several corresponding professional analytical agents in parallel based on enterprise diagnostic requests;

[0035] In this embodiment, time-series unified processing refers to standardizing data from different data sources with different time granularities and time identifiers to ensure consistency in the time dimension; the analysis dataset refers to multi-dimensional operational data after time-series unified processing, whose structure and format are standardized and can be directly used as input for professional analytical agents; the professional analytical agent refers to an automated module that performs data analysis and problem diagnosis for specific professional areas of enterprise operations, such as financial analysis, market analysis, supply chain analysis, and human resource analysis. Each professional analytical agent can have an independent analysis model and knowledge base.

[0036] S30: Synchronously input the analysis dataset into the professional analysis agent to obtain several diagnostic scores, including quantitative indicators, problem localization, and qualitative evaluation.

[0037] In this embodiment, the diagnostic score refers to the diagnostic results output by a single specialized analytical agent for its assigned domain. This score typically includes quantified indicator data, the specific location of the identified problem, and a qualitative assessment of the problem.

[0038] S40: Based on the preset integration logic, the obtained diagnostic scores are fused and conflict resolution is performed to generate a comprehensive diagnostic conclusion, which includes core problem attribution, overall health score and priority improvement item list;

[0039] In this embodiment, the integration logic refers to the preset rules, algorithms, or models used to fuse multiple diagnostic sub-results, resolve conflicts, and generate a comprehensive diagnostic conclusion. The comprehensive diagnostic conclusion refers to the final diagnostic result formed after fusing all diagnostic sub-results under the action of the integration logic. It can reflect the enterprise's operational status from multiple perspectives and can also provide actionable suggestions. The comprehensive diagnostic conclusion includes core problem attribution, overall health score, and a priority improvement item list. Core problem attribution refers to identifying and pointing out the root causes or key factors leading to the enterprise's operational problems in the comprehensive diagnostic conclusion. The overall health score is a comprehensive score that quantitatively assesses the enterprise's current operational status, reflecting the enterprise's overall performance in multiple dimensions. The priority improvement item list refers to improvement measures or action suggestions with a clear priority order proposed for the enterprise based on the comprehensive diagnostic conclusion.

[0040] S50: Based on the comprehensive diagnostic conclusion, match the corresponding text description template, data visualization chart component, and conclusion suggestion corpus from the pre-set report knowledge base;

[0041] In this embodiment, the report knowledge base refers to a collection of templates, components, and corpora used to generate enterprise operation diagnostic reports. The report knowledge base includes text description templates, data visualization chart components, and conclusion / suggestion corpora. Text description templates are preset structures and expression frameworks used to construct the text portion of the report, containing fillable placeholders for embedding specific information from the diagnostic conclusions. Data visualization chart components are preset modules used to graphically present quantitative data and logical relationships, such as bar charts, line charts, and pie charts. Conclusion / suggestion corpora are preset sets of text fragments or phrases used to generate the conclusions and suggestions sections of the report, which can be matched and combined based on the diagnostic results.

[0042] S60: By integrating the report with the intelligent agent, the comprehensive diagnostic conclusions are populated and associated with the selected text narrative templates, data visualization chart components, and conclusion suggestion corpus to generate an enterprise operation diagnostic report.

[0043] In this embodiment, the report integration agent refers to an automated module used to populate, associate, and render the comprehensive diagnostic conclusions with the templates, components, and corpora matched in the report knowledge base, and finally generate a structurally complete enterprise operation diagnostic report; the enterprise operation diagnostic report refers to the final generated structured report that reflects the enterprise's operational status and includes diagnostic conclusions, problem attribution, health score, and improvement suggestions.

[0044] Specifically, upon receiving a user-submitted enterprise diagnostic request, the request includes the target enterprise's identification information and the desired diagnostic period. Based on this identification information, various operational data of the enterprise within the specified period are obtained through a preset data connection method, such as directly exporting data files from the target enterprise's database. This operational data may include financial statements, sales records, inventory data, etc., collected in its original format. Further, after obtaining the raw multidimensional operational data, it is preprocessed, for example, by checking and adjusting the timestamps of different data sources to ensure consistency in the time dimension of all data, and organizing it into a unified table or file format to form an analysis dataset. Simultaneously, based on the diagnostic domain explicitly specified in the enterprise diagnostic request, pre-configured corresponding analytical agents are selected and activated. These analytical agents are typically capable of independent operation and are ready to receive the analysis dataset. Finally, each specialized analytical agent independently analyzes the received data and, based on its internal analytical model and... The system generates diagnostic scores for each domain based on rules. Further, it integrates and resolves conflicts in the obtained diagnostic scores according to pre-defined integration logic, generating a comprehensive diagnostic conclusion. This comprehensive diagnostic conclusion includes core problem attribution, overall health score, and a list of priority improvement items. After receiving diagnostic scores from all professional analysis agents, it resolves and summarizes conflicts using pre-defined integration rules, analyzes the summarized information, identifies the most significant problem as the core problem attribution, compiles a list of priority improvement items based on the severity and scope of the problem, calculates the overall health score, and constitutes the comprehensive diagnostic conclusion. Further, after the comprehensive diagnostic conclusion is generated, it searches a pre-defined report knowledge base for relevant text description templates, display chart components, and relevant conclusion suggestion corpora based on its main content. Finally, the report integration agent receives the comprehensive diagnostic conclusion and the matched report elements, integrates the specific numerical and textual information in the comprehensive diagnostic conclusion according to its pre-defined process logic, and forms an enterprise operation diagnostic report.

[0045] For example, suppose User A is a manager of a manufacturing company who wants a comprehensive diagnosis of the company's operations over the past year and to obtain a report containing improvement suggestions. User A submits a diagnostic request through the system, specifying the target company as "Manufacturing Company B," the diagnostic timeframe as "the past year," and focusing specifically on the business areas of "financial status" and "production efficiency." Upon receiving the diagnostic request, the system first obtains multi-dimensional operational data from Manufacturing Company B. Specifically, it exports the past year's financial transaction data, production record data, and human resource data from Company B's internal database, and simultaneously obtains the past year's market trends for the industry from publicly available external channels. The raw data acquired, including data from upstream and downstream of the supply chain, may have different time granularities and formats. Further, multi-dimensional operational data undergoes time-series unified processing to generate an analytical dataset. For example, the time granularity of all data is uniformly converted to "monthly" and aligned to the beginning of each month. Subsequently, based on the "financial status" and "production efficiency" domains specified in User A's diagnostic request, pre-defined financial analysis agents and production efficiency analysis agents are invoked in parallel. Further, the analytical dataset is synchronously input into the invoked financial analysis agents and production efficiency analysis agents. The financial analysis agent analyzes the financial transaction data and outputs "gross profit margin decreased by 5%" and "accounts receivable..." The system identifies quantitative indicators such as "increased turnover days by 10 days" and pinpoints the issue of "tight cash flow," commenting on "ineffective cost control." The production efficiency analysis agent analyzes production record data, outputting quantitative indicators such as "equipment utilization rate decreased by 8%" and "unit product energy consumption increased by 3%," and identifies the issue of "production line bottleneck," commenting on "insufficient equipment maintenance." This yields two independent diagnostic scores. Furthermore, based on pre-set integration logic, the two diagnostic scores are merged and conflict resolved. For example, the integration logic might identify a correlation between the "ineffective cost control" pointed out by the financial agent and the "increased unit product energy consumption" pointed out by the production agent, and resolve the conflict through pre-set rules. This information is then integrated to generate a comprehensive diagnostic conclusion. This conclusion may indicate that the core problem is attributed to "rising production costs leading to squeezed profit margins," giving an overall health score of 65, indicating a below-average level, and generating a list of priority improvement items, such as "optimizing production processes" and "strengthening equipment maintenance." Furthermore, based on the comprehensive diagnostic conclusion, corresponding text description templates, data visualization chart components, and conclusion suggestion corpora are matched from a pre-built report knowledge base. For example, a text template for comprehensive analysis of enterprise operation status, a profit trend chart component, an equipment utilization rate bar chart component, and suggestion corpora on "cost optimization" and "production efficiency improvement" are matched.Finally, the report integration agent populates and associates the comprehensive diagnostic conclusions with the selected text description templates, data visualization chart components, and conclusion suggestion corpora to generate an enterprise operation diagnostic report. For example, the report integration agent automatically populates information such as core problem attribution and overall health score into the text template. Simultaneously, it binds relevant financial and production data to matching chart components, automatically generating profit trend charts and equipment utilization bar charts. Combined with a priority improvement item list, it generates specific improvement suggestions from the suggestion corpus, forming a complete enterprise operation diagnostic report, which is then sent to the client used by user A.

[0046] Through the above process, User A does not need to wait for the expert team's analysis, and avoids the limitations of a single automated analysis dimension. At the same time, the diagnostic results and report generation are closely linked, enabling rapid and standardized enterprise operation diagnosis and report output.

[0047] Compared to the traditional approach that relies on expert teams for manual analysis, this method achieves multi-faceted and synchronized analysis of enterprise operational data by calling multiple specialized analytical agents in parallel. In the example above, the financial analysis agent and the production efficiency analysis agent can simultaneously analyze data in their respective fields, thus avoiding the problems of time-consuming manual analysis, reliance on personal experience, and inconsistent evaluation standards. As a result, the efficiency and objectivity of the analysis and diagnosis process are improved, and the comprehensiveness of the diagnostic conclusions is enhanced. At the same time, by having each specialized agent independently analyze and output diagnostic sub-results, and then integrating and resolving conflicts through integration logic, the depth and breadth of the comprehensive diagnostic conclusions can be further improved.

[0048] Furthermore, this method effectively solves the problem of the disconnect between the diagnosis and report generation stages. In the example above, after the comprehensive diagnostic conclusion is generated, the text description template, data visualization chart components, and conclusion suggestion corpus are matched from the report knowledge base based on the comprehensive diagnostic conclusion. The data is then filled and associated through the report integration agent, and finally a complete enterprise operation diagnostic report is generated. This shortens the overall cycle from diagnosis to report output, reduces the workload of manual sorting, chart creation, and text compilation, and reduces the risk that the report quality will be affected by the subjective factors of the writer. As a result, it can quickly and accurately respond to the dynamic diagnostic needs of enterprises.

[0049] In summary, this embodiment provides an efficient, comprehensive, objective, and standardized enterprise operation diagnosis and report generation solution by constructing a multi-agent collaborative diagnostic framework and realizing the linkage between diagnostic conclusions and report generation. It can effectively solve the technical problems of low efficiency, limited analysis, and disconnect between report generation and existing technologies.

[0050] In one embodiment, such as Figure 3As shown, step S10 includes:

[0051] S11: Analyze enterprise diagnostic requests to determine the business area and time frame;

[0052] In this embodiment, step S11 aims to clarify the specific requirements of this diagnosis. For example, natural language processing technology can be used to perform semantic analysis on the unstructured diagnosis request text to extract business domain keywords and time range information; or, when the diagnosis request is submitted in the form of a structured form, preset fields can be read directly to determine the business domain and time range.

[0053] S12: Based on the defined business area and time range, collect first operational data from the business system associated with the target enterprise through a preset data interface. The first operational data includes financial transaction data, production record data, and human resource data.

[0054] In this embodiment, the preset data interface may include real-time data interaction with internal enterprise systems such as ERP, CRM, and MES via API; or, directly accessing the internal enterprise database through a database connector to execute SQL queries to extract the required data.

[0055] S13: Collect second operational data from external data sources associated with the target enterprise through the data collection terminal. The second operational data includes industry market data, public data, and upstream and downstream supply chain data.

[0056] In this embodiment, the data acquisition end can be a dedicated web crawler module used to periodically crawl publicly available data such as industry reports, news information, and competitors' public financial reports; or it can be a data integration channel established with a third-party data service provider to obtain macroeconomic indicators, industry prosperity indices, and operating data of upstream and downstream enterprises in the supply chain through subscription or purchase.

[0057] S14: Perform data cleaning and field standardization on the first and second operational data to form multidimensional operational data.

[0058] In this embodiment, data cleaning may include identifying and processing missing values ​​(e.g., filling in with the mean, interpolating, or deleting records), correcting erroneous data (e.g., through rule validation or outlier detection), and removing duplicate records. Field standardization is used to unify the data format, units, and naming conventions of different data sources. For example, different currency units are uniformly converted to standard currency units, date formats are uniformly standardized to "YYYY-MM-DD", and employee numbers and work numbers in different systems are uniformly mapped to employee IDs, thereby ensuring the consistency of all data in structure and semantics, facilitating integration and analysis.

[0059] Specifically, the process begins by parsing the enterprise's diagnostic requests to clarify the business areas and timeframes for the diagnosis. Based on this, targeted collection of primary and secondary operational data is conducted from the enterprise's internal business systems and diverse external data sources. This ensures the comprehensiveness and multi-dimensionality of the data required for the diagnosis, covering both internal operational details and external market environment considerations. The collected heterogeneous data undergoes data cleaning and field standardization to eliminate noise, inconsistencies, and redundancy, transforming the raw data into structured, multi-dimensional operational data. This provides professional analytical agents with analytical datasets that can be directly used for analysis, thereby improving the starting data quality of the entire enterprise diagnostic process.

[0060] For example, suppose a manufacturing company initiates a business diagnostic request to assess its production and operational efficiency over the past six months. First, the request is parsed to identify the business area as production and operations, with a timeframe of the past six months. Based on this, first-level operational data is collected from the company's MES system via a pre-defined data interface, such as production order completion rate, equipment utilization rate, defect rate, and production cycle data. Simultaneously, production-related financial transaction data, such as manufacturing costs and inventory turnover rate, is collected from its financial system. Human resource data for the production department, such as employee hours and training records, is also collected from its human resources system. Furthermore, second-level operational data is collected from external data sources, such as subscribed industry reports, via a data acquisition terminal. The system acquires industry market data such as average production efficiency and product quality complaint rates from peers; it also obtains publicly available data such as raw material price indices and labor cost changes from the National Bureau of Statistics website; and it acquires supply chain data such as on-time delivery rates of upstream suppliers and order fluctuations of downstream customers from the supply chain management platform. After all data collection is completed, the collected data is cleaned, for example, by removing abnormal equipment utilization rate readings caused by sensor failures in the MES system, filling in certain missing cost items in the financial system, and standardizing fields, such as unifying all production cycle units to "hours" and mapping product codes from different systems to standard codes, ensuring consistency in format and semantics of all data, ultimately forming multi-dimensional operational data.

[0061] Through the above technical solutions, this application can achieve a precise understanding of enterprise diagnostic requests, thereby obtaining comprehensive and highly relevant data from both internal and external sources. Combined with data cleaning and field standardization, it can effectively solve the problems of scattered data sources, heterogeneous formats, and inconsistent quality, providing a data foundation for professional analytical agents, thereby improving the accuracy and efficiency of agent analysis, avoiding diagnostic biases caused by data quality issues, and enhancing the reliability of the final comprehensive diagnostic conclusions and enterprise operation diagnostic reports.

[0062] In one embodiment, such as Figure 4 As shown, step S20 includes:

[0063] S21: Identify the original time stamp and data frequency corresponding to each data item in the multidimensional operational data;

[0064] In this embodiment, step S21 aims to clarify the time attribute of each data item in the multidimensional operational data; wherein, the original time identifier refers to the original timestamp of the data record, such as the transaction time, production completion time, etc.; the data frequency refers to the data update or collection cycle, such as daily, weekly, monthly, quarterly, etc.; the implementation may include: obtaining the time identifier and frequency information by parsing the metadata or data dictionary of the data source; or, inferring the data frequency by performing statistical analysis on the data sample, such as calculating the interval distribution of adjacent timestamps.

[0065] S22: Convert data items with different original time stamps and data frequencies to a preset standard diagnostic time granularity and align them to a preset time series reference point to obtain the analysis dataset;

[0066] In this embodiment, step S22 aims to standardize heterogeneous time-series data to maintain consistency across the time dimension. The preset standard diagnostic time granularity can be daily, weekly, monthly, quarterly, etc., depending on the granularity requirements of the diagnostic request and the needs of business analysis. Aligning to a preset time series reference point means unifying the starting point of all data's time series, for example, to the beginning of each month or the start of each quarter. This can be achieved by: downsampling high-frequency data to the standard granularity through aggregation; and upsampling low-frequency data to the standard granularity through interpolation or forward padding. Simultaneously, time-series processing algorithms are used to ensure that all data points are aligned on the time axis.

[0067] S23: Parse the enterprise diagnostic request, determine several target diagnostic dimensions, and match and call the corresponding professional analytical agent from the predefined set of professional analytical agents based on the determined target diagnostic dimensions.

[0068] In this embodiment, step S23 aims to understand user needs and dynamically select the most suitable analysis tool accordingly. Enterprise diagnostic requests are typically given in the form of natural language or structured queries, containing descriptions of specific business areas, timeframes, and key metrics. The purpose of parsing these requests is to extract the core diagnostic intent, i.e., the target diagnostic dimension, such as financial health, market competitiveness, or supply chain resilience. This can be achieved by: using natural language processing technology to perform semantic analysis on the request text, extracting keywords and phrases, and mapping them to a predefined diagnostic dimension ontology; or by using a rule engine or machine learning model to automatically identify and recommend target diagnostic dimensions based on specific patterns or historical diagnostic cases in the request. After determining the target diagnostic dimension, from a predefined set of professional analytical agents, one or more analytical agents that best match the target diagnostic dimension are matched and invoked based on the analytical areas and functions that the analytical agents excel at. For example, a financial analytical agent is invoked for financial health, and a market analytical agent is invoked for market competitiveness.

[0069] Specifically, each data item in the multidimensional operational data is analyzed to identify its inherent original time signature and data frequency to understand the data's temporal characteristics. Based on this, data items with different time attributes are uniformly converted to a preset standard diagnostic time granularity and aligned to a preset time series benchmark. This effectively eliminates inconsistencies in time dimensions caused by diverse data sources, generating a consistent analytical dataset that can be directly used for analysis. Simultaneously, enterprise diagnostic requests are parsed to identify several core target diagnostic dimensions, avoiding the ambiguity and inefficiency of manual judgment and making diagnostic objectives clearer. Furthermore, based on the target diagnostic dimensions, the most suitable professional analytical agent is matched and invoked from a predefined set of professional analytical agents to perform the corresponding dimension analysis, ensuring a high degree of synergy between the generation of the analytical dataset and the invocation of professional analytical agents.

[0070] Through the above technical solutions, this application can effectively solve the problem of heterogeneity in the time dimension of multidimensional operational data, ensuring that all data have a unified time granularity and alignment benchmark before analysis, thereby improving the accuracy and reliability of data analysis. At the same time, by parsing enterprise diagnostic requests, identifying diagnostic needs, and matching and calling the most suitable analytical agent accordingly, unnecessary waste of computing resources can be avoided, and the pertinence and depth of diagnostic analysis can be ensured. Through the combination of unified data time-series processing and intelligent agent calling, the enterprise diagnostic process can be made more intelligent, improving diagnostic efficiency and the quality of diagnostic results.

[0071] In one embodiment, step S23 includes:

[0072] S231: Perform semantic recognition on the enterprise diagnosis request and extract business domain keywords and diagnosis description keywords;

[0073] In this embodiment, semantic recognition of enterprise diagnostic requests aims to understand the deeper meaning and intent of these requests through technical means, thereby transforming unstructured natural language requests into processable structured information. This can be achieved using rule-based methods, such as matching and syntactic analysis using a pre-defined keyword list; or using machine learning methods, such as using deep learning models to extract features and classify intents from the request text. Through semantic recognition, business domain keywords, such as financial status, market expansion, and supply chain efficiency, as well as diagnostic description keywords, such as declining profits, excessive inventory, and customer churn, can be accurately extracted.

[0074] S232: Map and match the extracted business domain keywords with the preset diagnostic knowledge base, and combine them with diagnostic description keywords to determine several target diagnostic dimensions and their invocation priorities;

[0075] In this embodiment, the diagnostic knowledge base refers to a structured collection of information that includes the relationships between different business domains, diagnostic issues, and corresponding diagnostic dimensions. The diagnostic knowledge base can be an ontology model defining concepts, attributes, and relationships; or it can be a rule base containing logical judgments. Mapping matching refers to comparing and associating extracted keywords with entries in the knowledge base. Methods such as exact matching or semantic similarity calculation can be used. Through mapping matching, abstract keywords can be transformed into specific, actionable target diagnostic dimensions, such as "profitability analysis" or "operational efficiency assessment." Simultaneously, by combining detailed information about the diagnostic description keywords, the calling priority of each target diagnostic dimension can be evaluated and determined. This calling priority can be dynamically calculated based on the frequency of keyword occurrence, the strength of association with core issues, or preset business rules, ensuring that the diagnostic process focuses on the most critical or urgent issues.

[0076] S233: Based on the determined target diagnostic dimensions and their invocation priorities, the corresponding professional analytical agents are invoked sequentially from the predefined set of professional analytical agents.

[0077] In this embodiment, the professional analytical agent set is a collection library containing multiple independent agent modules that focus on specific diagnostic dimensions or analytical tasks. Each analytical agent has the ability to process specific types of data and execute specific analytical algorithms. Sequential invocation refers to activating and running the corresponding professional analytical agents one by one or in batches according to a pre-determined invocation priority order. This sequential invocation mechanism helps optimize resource allocation, ensures that core diagnostic dimensions are processed first, and allows the analysis results of subsequent analytical agents to be adjusted or refined based on the output of previous analytical agents.

[0078] It should be noted that the professional analysis agent in this embodiment refers to a software entity with autonomous processing capabilities built to solve analysis problems in a specific business domain; each professional analysis agent is not a single algorithm, but an integrated functional module that encapsulates domain knowledge, data analysis models and logical judgment rules.

[0079] For example, each specialized analytical agent may include a data input interface, a core processing engine, a domain knowledge base or model library, and a result output interface. The data input interface is configured to receive an analytical dataset that has undergone time-series unified processing, capable of identifying and extracting data subsets relevant to the analytical agent's domain. For instance, the input interface of a financial analytical agent is configured to identify fields related to financial indicators. The specific form of the core processing engine depends on the domain problem being solved. It can be a prediction or classification model trained on historical data, such as a neural network or gradient boosting decision tree; it can also be a state machine or inference engine based on expert rules and logic; or it can be a graph analysis engine that traverses, queries, and computes a knowledge graph. The core processing engine drives the entire analysis process. The domain knowledge base or model library stores static or dynamic knowledge supporting the operation of the core processing engine. For model-based engines, the domain knowledge base or model library can store model parameters; for rule-based or knowledge graph-based engines, the domain knowledge base or model library can store specific business rules, industry standard thresholds, entity relationship networks, etc. For example, the knowledge base of the production cost analysis agent stores standard energy consumption ranges and yield rate benchmarks for each production stage. The result output interface is configured to format the analysis results of the core processing engine into a unified diagnostic sub-result data structure and output it. This diagnostic sub-result data structure includes at least three standardized fields: quantitative indicators, problem localization, and qualitative evaluation, ensuring that the downstream integration logic can parse it consistently. In summary, in this embodiment, multiple professional analysis agents are deployed in parallel, accepting calls through a unified API or message protocol. Each professional analysis agent works independently during the analysis phase and collaborates through standardized output.

[0080] Meanwhile, other architectural approaches for professional analytical agents can be modified and configured according to actual scenarios and needs, and will not be elaborated here.

[0081] Specifically, the system extracts key business domain keywords and diagnostic description keywords from unstructured request text. These keywords form the basis for understanding the user's diagnostic intent. Subsequently, the system maps and matches these extracted keywords with a pre-defined diagnostic knowledge base. This knowledge base contains rich domain experience and diagnostic logic, transforming abstract keywords into concrete, actionable target diagnostic dimensions. During this process, by combining refined information from the diagnostic description keywords, the system can further evaluate and determine the invocation priority of each target diagnostic dimension, ensuring that the diagnostic process focuses on the most critical and urgent issues. Finally, based on these determined target diagnostic dimensions and their invocation priorities, the system sequentially invokes the appropriate agents from a predefined set of professional analytical agents according to priority. This mechanism ensures the intelligence and customization of the diagnostic process, avoiding blind invocation or omission of key analytical steps, enabling professional analytical agents to process and analyze datasets in a targeted manner, thereby improving the accuracy and efficiency of the diagnosis. In this way, this application achieves a deeper understanding and more refined response to enterprise diagnostic requests, resulting in more accurate and effective generation of subsequent diagnostic results.

[0082] Through the above technical solution, this application can parse enterprise diagnostic requests, accurately extract business domain keywords and diagnostic description keywords from fuzzy natural language descriptions, thereby determining relevant target diagnostic dimensions based on a preset diagnostic knowledge base and assigning them reasonable invocation priorities. This ensures that professional analytical agents can be invoked in a targeted manner, avoiding unnecessary waste of computing resources and allowing the diagnostic process to focus on the enterprise's most core or urgent issues. Therefore, compared to determining target diagnostic dimensions solely based on diagnostic requests, the solution of this application can significantly improve the accuracy and efficiency of diagnosis, ensuring that the generated diagnostic results are more targeted.

[0083] In one embodiment, step S40 includes:

[0084] S41: Based on the multi-dimensional evaluation system and related knowledge network associated with the pre-set integration logic, the obtained diagnostic scores are structured and semantically aligned to obtain several aligned scores.

[0085] In this embodiment, the multi-dimensional evaluation system and associated knowledge network associated with the pre-defined integration logic form the basis for achieving diagnostic result fusion and conflict resolution. The multi-dimensional evaluation system can define various aspects of enterprise operations, such as finance, marketing, production, and human resources, and includes quantitative indicators, measurement standards, weighting rules, and health level classifications for each dimension. It can be represented as a predefined rule set, an expert knowledge base, or a model trained based on machine learning. The associated knowledge network stores deep knowledge such as causal relationships, influence paths, and best practices within the enterprise operation domain, and can be implemented using graph databases, ontological models, or knowledge graphs constructed based on natural language processing. The obtained diagnostic results are then processed. The segmentation results undergo structured parsing and semantic alignment to transform heterogeneous diagnostic information from different professional analytical agents into a unified and processable format. Structured parsing converts unstructured or semi-structured qualitative assessments into machine-understandable structured data, for example, through natural language processing techniques such as named entity recognition, relation extraction, and sentiment analysis. Semantic alignment maps information of different terms or granularities that may be used by different agents to a predefined unified semantic representation in a multidimensional evaluation system and associated knowledge network, for example, through ontology mapping, word vector similarity calculation, or rule matching. After the above processing, several aligned segmentation results with consistent semantics and structure can be obtained.

[0086] S42: Based on the enterprise's diagnostic request, obtain the corresponding historical diagnostic conclusions, map the quantitative indicators in the alignment results to a unified measurement space in the multidimensional evaluation system, and calibrate the mapping results based on the feedback data of the historical diagnostic conclusions.

[0087] In this embodiment, historical diagnostic conclusions may include past diagnostic reports, improvement suggestions, and feedback data on the effects of their implementation on the target enterprise or similar enterprises; the unified metric space is a predefined standard scale in the multidimensional evaluation system, used to normalize or standardize quantitative indicators of different dimensions to make them comparable; the mapping results are calibrated based on the feedback data of historical diagnostic conclusions, which means that the interpretation and mapping values ​​of the current quantitative indicators can be dynamically adjusted according to the actual effects of historical improvement measures, for example, through regression analysis, reinforcement learning, or expert rules, so that the diagnostic results are closer to the actual operating conditions.

[0088] S43: Infer the coupling relationship and influence strength among the diagnostic dimensions in the current diagnostic context through the associated knowledge network, and assign dynamic contribution weights to the calibrated alignment results based on the coupling relationship and influence strength.

[0089] In this embodiment, coupling relationship and influence strength describe the degree and direction of interaction between different diagnostic dimensions in the associated knowledge network, which can be achieved through graph theory algorithms, causal inference models, or expert experience rule deduction; dynamic contribution weight is a variable weight assigned to each alignment result based on the focus of the current enterprise diagnostic request, specific industry characteristics, and coupling relationship between each diagnostic dimension, for example, through machine learning models, rule engines, or dynamic programming algorithms, to ensure that more important diagnostic information can play a greater role in the comprehensive evaluation.

[0090] S44: Identify the points of disagreement in the qualitative evaluation and problem localization of each alignment result, and based on the causal relationship and dynamic contribution weight revealed by the associated knowledge network, perform evidence fusion and collaborative reasoning on the points of disagreement to generate a consistent judgment.

[0091] In this embodiment, identifying the points of disagreement in the qualitative assessment and problem localization of each alignment result, and then performing evidence fusion and collaborative reasoning on these points based on the causal relationships and dynamic contribution weights revealed by the associated knowledge network to generate a consistent judgment, is a key step in resolving conflicts in multi-agent diagnostic results. Points of disagreement refer to inconsistencies or contradictions in the qualitative descriptions or problem attributions of different agents, which can be identified through text similarity comparison, keyword conflict detection, or logical rule conflict detection. Evidence fusion and collaborative reasoning combine the causal relationships and dynamic contribution weights in the associated knowledge network to comprehensively judge these disagreements, for example, by using Bayesian networks, DS evidence theory, or multi-source information fusion algorithms, thereby arriving at more reliable and persuasive conclusions and forming a consistent judgment.

[0092] S45: Based on consistency judgment, an overall health score is generated through a pre-set synthesis function in a multi-dimensional evaluation system. The core problem attribution is located by tracing the key path in the related knowledge network, and a priority improvement item list is generated by combining dynamic contribution weights with a pre-set repair cost assessment model.

[0093] In this embodiment, the synthesis function is a pre-set mathematical model in the multidimensional evaluation system, used to aggregate multiple quantitative indicators into an overall health score, such as a weighted average, fuzzy comprehensive evaluation, or multi-objective optimization function, which can be determined based on the actual scenario and needs; the critical path is the path from the core problem to its root cause or main influencing factor in the associated knowledge network, which can be identified through graph traversal algorithms, causal chain analysis, or influence path analysis; the repair cost assessment model is a pre-set model used to assess the resources required to solve a specific problem, such as based on historical data statistics, expert experience rules, or machine learning prediction models. By integrating the above information, a list of priority improvement items with guiding significance can be generated.

[0094] Specifically, the solution proposed in this application effectively solves the problem of integrating heterogeneous information by introducing a multi-dimensional evaluation system and an associated knowledge network to perform structured analysis and semantic alignment of diagnostic scores from different professional analytical agents. It calibrates quantitative indicators using feedback data from historical diagnostic conclusions and infers dynamic contribution weights based on the associated knowledge network, making the diagnostic results more context-adaptive and accurate. Crucially, by identifying and resolving disagreements in the diagnostic scores and performing evidence fusion and collaborative reasoning, the consistency and reliability of the final comprehensive diagnostic conclusion can be improved. Through this comprehensive integration and conflict resolution mechanism, it is possible to understand the enterprise's operational status from multiple perspectives and generate highly accurate comprehensive diagnostic conclusions, including clear core problem attributions, quantified overall health scores, and a clear list of priority improvement items.

[0095] For example, suppose the purpose of a company's diagnostic request is to assess the operational status of a manufacturing company, and corresponding diagnostic scores are obtained from multiple professional analytical agents such as finance, production, and marketing. These diagnostic scores include "declining profit margin" reported by the financial agent, "increased equipment failure rate" reported by the production agent, and "slightly decreased customer satisfaction" reported by the marketing agent. To integrate this information, a pre-defined multi-dimensional evaluation system is used to classify profit margin as a financial health indicator, equipment failure rate as a production and operation efficiency indicator, and customer satisfaction as a market competitiveness indicator. At the same time, the associated knowledge network stores causal relationships such as "an increase in equipment failure rate will lead to an increase in production costs, which in turn will affect profit margin" and "product quality problems will reduce customer satisfaction and may be related to equipment failure."

[0096] Furthermore, the diagnostic results are structured and semantically aligned, mapping quantitative indicators such as "declining profit margin" to a unified measurement space, and calibrating the mapping results based on historical diagnostic conclusions. At the same time, depending on the focus of the enterprise's diagnostic request, for example, if the request focuses particularly on production efficiency, the coupling relationship between various diagnostic dimensions is deduced through the association knowledge network, and higher dynamic contribution weights are assigned to production-related diagnostic results.

[0097] Furthermore, during the processing, potential discrepancies may be identified between "increased equipment failure rate" and "slightly decreased customer satisfaction," as they may point to different root causes. In this case, based on the causal relationship in the knowledge network that "equipment failure leads to unstable product quality, thus affecting customer satisfaction," and combined with the quantitative indicators and dynamic contribution weights of each diagnostic score, evidence fusion and collaborative reasoning are performed. For example, if the quantitative indicator of equipment failure rate is abnormally significant and indirectly mentioned by multiple analytical agents, it may be determined that equipment failure is a deeper cause of decreased customer satisfaction, thus generating a consistent judgment. Based on the generated consistent judgment, the calibrated and weighted indicators are aggregated into an overall health score, such as 75 points, through a pre-set synthesis function in the multi-dimensional evaluation system. Simultaneously, the critical path in the knowledge network is traced, for example, from "increased equipment failure rate" to "insufficient equipment maintenance" or "equipment aging," thereby locating the core problem attribution. Finally, by integrating the dynamic contribution weights, the pre-set repair cost assessment model, and the impact propagation range assessment value of the critical path, a priority improvement item list is generated, such as "prioritize strengthening equipment maintenance, and secondly consider introducing new production equipment."

[0098] Through the above technical solutions, this application can effectively integrate heterogeneous diagnostic information from multiple professional analytical agents and resolve potential conflicts and inconsistencies. Specifically, by introducing a multi-dimensional evaluation system and a related knowledge network, it can achieve structured analysis, semantic alignment, dynamic calibration, and contextual weighting of diagnostic results. More importantly, through evidence fusion and collaborative reasoning mechanisms, this application can effectively identify and resolve discrepancies in diagnostic results, thereby generating reliable and insightful comprehensive diagnostic conclusions. This enables enterprises to obtain more accurate overall health assessments, clearer attributions of core issues, and more actionable lists of priority improvement items, thereby improving the quality and practicality of enterprise diagnostic reports.

[0099] In one embodiment, step S45 includes:

[0100] S451: Based on consistency judgment, extract the quantitative indicators of each aligned score after calibration and dynamic contribution weighting;

[0101] In this embodiment, step S451 aims to filter key quantitative data for scoring and attribution from the pre-processed and fused diagnostic scores. Consistency judgment ensures that the extracted quantitative indicators have consensus or resolved conflicts among different professional analytical agents, thereby improving data reliability. The calibration process corrects deviations in the quantitative indicators, while dynamic contribution weights reflect the importance of each indicator in the current diagnostic context, making the extracted indicators more representative and accurate. This can be achieved by a preset data filtering module selecting the corresponding quantitative indicator fields from the data structure of the aligned scores based on the consistency judgment result, and applying pre-calculated calibration factors and dynamic contribution weights for numerical adjustment. Alternatively, a preset data processing engine performs data query and calculation operations upon receiving the consistency judgment signal, extracting and weighting relevant quantitative indicators from the stored aligned scores.

[0102] S452: Input the extracted quantitative indicators into the pre-set synthesis function of the multidimensional evaluation system for weighted aggregation, and adjust the confidence of the weighted aggregation result in combination with dynamic contribution weights to generate an overall health score;

[0103] In this embodiment, step S452 aims to utilize a structured evaluation model to integrate discrete quantitative indicators into a health score that reflects the overall operational status of the enterprise. The synthesis function defines the logic of how different indicators are combined to form the final score, weighted aggregation considers the importance of each indicator, and confidence adjustment further enhances the reliability of the score, reflecting the degree of certainty in the diagnostic results. For example, a linear weighted summation function, a nonlinear mapping function, or a prediction function based on a machine learning model can be used as the synthesis function, and the specific settings can be tailored to the actual scenario and requirements. Confidence adjustment can be achieved based on the distribution of dynamic contribution weights, the strength of consistency judgments, or the accuracy of historical diagnoses.

[0104] S453: Identify key issues in consistency judgment, traverse the related knowledge network starting from the key issues, and trace the lower-level related nodes that lead to the key issues;

[0105] In this embodiment, step S453 aims to delve deeper from macroscopic diagnostic conclusions to specific, actionable root causes of problems. Key issues are central to the diagnostic process, while the related knowledge network provides structured relationships between problems and between problems and their causes. By traversing the related knowledge network, one can start from the surface-level problems and delve deeper to reveal their underlying causes. For example, key issues can be extracted from textual descriptions of consistency judgments using natural language processing techniques, or obtained through predefined rule matching. The related knowledge network can be a graph database, where nodes represent concepts, entities, or problems, and edges represent relationships between them, such as causality, inclusion, and influence. The traversal process can employ depth-first search or breadth-first search algorithms.

[0106] S454: Identify the impact intensity of each path based on the traced lower-level related nodes, determine the path with the largest cumulative impact intensity as the critical path, and locate the core problem attribution based on the critical path;

[0107] In this embodiment, step S454 aims to further refine the accuracy of problem attribution by quantifying the impact of different paths on key issues to identify the most important or fundamental causal path. The impact strength can be a preset weight, a correlation strength learned from historical data, or a value set based on expert knowledge. The cumulative impact strength reflects the comprehensive impact on the entire chain from the root cause to the manifestation. By identifying the critical path, the core issues causing enterprise problems can be accurately identified. For example, the impact strength can be stored as an attribute of the edges in the correlation knowledge network, and the cumulative impact strength can be calculated by multiplying or summing the weights of all edges on the path. After determining the critical path, the starting node or key intermediate node on the path can be located as the core problem attribution.

[0108] S455: Generate a list of priority improvement items by combining the dynamic contribution weight, the cost assessment value output by the repair cost assessment model, and the impact propagation range assessment value of the critical path.

[0109] In this embodiment, step S455 aims to transform the diagnostic results into action plans with practical guidance. The priority improvement item list not only identifies the problems requiring improvement but also ranks them according to their importance, implementation cost, and potential impact. Dynamic contribution weights reflect the importance of the problem, the repair cost assessment model provides an estimate of the resources required to implement improvement measures, and the impact propagation scope assessment value measures the positive ripple effects that solving the problem may bring. For example, priorities can be calculated using a comprehensive scoring formula that weights and combines the above three factors. The repair cost assessment model can be a predictive model based on historical data and expert experience, and the impact propagation scope assessment value can be obtained by analyzing the coverage and connectivity of the critical path in the associated knowledge network.

[0110] Specifically, in generating comprehensive diagnostic conclusions, to ensure their accuracy, precision, and practicality, the first step is to extract key quantitative indicators from the aligned scores after calibration and weighting, based on consistency judgment. These key quantitative indicators are data foundations with high consensus and reliability after multi-agent collaborative processing and conflict resolution. Subsequently, the extracted quantitative indicators are input into a pre-defined synthesis function of the multi-dimensional evaluation system for weighted aggregation, thereby generating an overall health score that comprehensively reflects the enterprise's operational status. During this process, the confidence level of the aggregation results is adjusted by combining dynamic contribution weights, which further improves the reliability of the score and avoids biases caused by data uncertainty or residual conflicts. Simultaneously, to delve into the root causes of problems, this solution identifies key issues revealed in the consistency judgment and uses this as a starting point for deep traversal of the associated knowledge network, tracing down to the lower-level relationships leading to the key issues. The connection nodes can construct a complete path from the apparent problem to the underlying cause. Based on this, the impact intensity along these paths is further identified, and the cumulative impact intensity is calculated to accurately determine the critical path with the greatest cumulative impact intensity. This critical path directly points to the core problem attribution in the enterprise's operations, enabling the diagnostic results to address the essence of the problem. Finally, to transform the diagnostic results into actionable improvement suggestions, this solution comprehensively considers dynamic contribution weights, the cost assessment value output by the repair cost assessment model, and the impact propagation scope assessment value of the critical path. The dynamic contribution weights reflect the importance of the problem, the repair cost assessment value provides economic considerations for implementing improvement measures, and the impact propagation scope assessment value quantifies the potential positive impact of solving the problem. Through the comprehensive evaluation of these multi-dimensional information, a list of priority improvement items can be generated, providing the enterprise with a clear and efficient action guide.

[0111] As a specific implementation method, when generating an overall health score, the process can begin by extracting, based on a consensus judgment, calibrated and dynamically weighted financial, operational, and market indicators from the aligned scores output by various professional analytical agents. For example, if the consensus judgment indicates a problem with a company's profitability, then quantitative indicators related to profitability will be extracted. Subsequently, the extracted quantitative indicators can be input into a pre-defined synthesis function. This synthesis function can be a multi-factor linear regression model based on expert experience or a neural network model trained on historical data. For instance, this synthesis function can weight and sum indicators such as profit margin and sales growth rate according to preset weights, and then adjust the confidence level of the final health score using dynamic contribution weights, resulting in an overall health score between 0 and 100. When identifying the root cause of core problems, key issues indicated in the consensus judgment can be identified, such as a continuously declining profit margin. Starting from this key issue, a pre-constructed knowledge network can be traversed. This knowledge network can be a graph database, where nodes represent various factors in company operations, such as product costs, sales strategies, market competition, and supply chain efficiency, and edges represent these factors. The causal or influencing relationships between factors are identified, along with their influence strength values. For example, starting from a point of continuously declining profit margins, one can trace down to lower-level related nodes such as excessively high product costs and declining sales. Furthermore, excessively high product costs can be traced back to deeper-level nodes such as rising raw material procurement prices and low production efficiency. The influence strength along these paths is then identified. For instance, if the influence strength of "rising raw material procurement prices" on "excessively high product costs" is 0.8, while the influence strength of "low production efficiency" is 0.6, then the former will be given a higher weight when calculating the cumulative influence strength. Finally, the factor with the highest cumulative influence strength is considered. The path is determined, for example, "rising raw material procurement prices, excessively high product costs, and continuously declining profit margins" are identified as critical paths, and "rising raw material procurement prices" are positioned as the core problem attribution. When generating a priority improvement item list, the above core problem attribution is comprehensively considered, combined with their dynamic contribution weight, the cost assessment value output by the repair cost assessment model, and the impact propagation scope assessment value of the critical path. Through a preset priority calculation formula, a sorted improvement item list is generated, for example: "1. Re-evaluate and negotiate raw material supplier contracts; 2. Optimize production processes to improve efficiency; 3. Explore alternative raw material solutions."

[0112] Through the above technical solutions, this application can significantly improve the accuracy, depth, and practicality of enterprise diagnostic conclusions. Specifically, by extracting calibrated and weighted quantitative indicators based on consistency judgments and adjusting the confidence level using the synthesis function of a multi-dimensional evaluation system, the generated overall health score will be more objective and reliable, thus avoiding the bias caused by a single indicator or simple aggregation. At the same time, by using a related knowledge network to deeply trace key issues and quantifying the impact intensity of each path to locate the core problem attribution, the diagnostic results can directly point to the fundamental problems of enterprise operations, rather than remaining at the surface phenomena, thereby enhancing the accuracy of problem attribution. In addition, by comprehensively considering dynamic contribution weights, repair costs, and the scope of impact propagation to generate a priority improvement item list, it can ensure that the provided improvement suggestions are not only highly relevant and important, but also more feasible and effective in implementation, thus providing enterprises with more instructive and actionable decision support.

[0113] In one embodiment, step S60 includes:

[0114] S61: Analyze the comprehensive diagnostic conclusions and extract key entities, quantitative values, and logical relationships;

[0115] In this embodiment, step S61 aims to transform the structured or semi-structured comprehensive diagnostic conclusion into information units that can be processed subsequently. Here, key entities refer to specific objects involved in the diagnostic conclusion, such as enterprise departments, products, markets, and time periods; quantified values ​​refer to various indicator data included in the diagnostic conclusion, such as growth rate, score, and cost; logical relationships refer to causal, parallel, or progressive associations between entities and values. One implementation method is to use natural language processing technology to identify key entities through a named entity recognition model, extract quantified values ​​through regular expressions or value extraction algorithms, and identify logical relationships through a relation extraction model or dependency parsing. Another implementation method is to predefine a structured template or ontology for the comprehensive diagnostic conclusion, map the diagnostic conclusion to this template or ontology, and thus automatically parse out key entities, quantified values, and logical relationships.

[0116] S62: Based on the extracted key entities, quantified values, and logical relationships, semantic adaptation and filling are performed with the selected text narrative template to generate narrative text paragraphs;

[0117] In this embodiment, step S62 aims to embed the parsed diagnostic information into a preset text narrative template to form a fluent and accurate report text. Semantic adaptation refers to ensuring that the filled content is semantically consistent and coherent with the template context, avoiding awkward splicing. One implementation method is to use a rule-based template engine, pre-setting placeholders in the template and filling these placeholders according to extracted key entities, quantified values, and logical relationships, while adjusting them in conjunction with predefined grammatical rules. Another implementation method is to utilize natural language generation technology, taking structured diagnostic information as input, and using a trained generation model combined with the style and structure of the selected text narrative template to generate a narrative text paragraph that conforms to semantics and grammar.

[0118] S63: Based on the characteristics of the core problem attribution and the priority improvement item list, match and instantiate the corresponding chart type from the selected data visualization chart component library, and bind the relevant quantitative values ​​and logical relationships to the chart data source to obtain the data visualization chart;

[0119] In this embodiment, step S63 aims to present the quantitative information in the diagnostic conclusion in an intuitive and easy-to-understand chart format. The core problem attribution and priority improvement item list typically includes information such as the changing trends, comparative analysis, or composition ratios of key indicators, which are suitable for chart display. One implementation method is to establish a mapping rule between chart types and data characteristics. For example, time series data is matched with line charts, categorical comparison data with bar charts, and proportional composition data with pie charts. Based on the data characteristics in the core problem attribution and priority improvement item list, the most suitable chart type is automatically matched, and a chart library is called for instantiation. Another implementation method is to utilize a machine learning model to analyze the correlation between chart usage patterns in historical reports and diagnostic conclusions, recommending and generating the most effective data visualization charts to express the current diagnostic conclusions, and using the extracted quantitative values ​​and logical relationships as data source input.

[0120] S64: Based on the order of the priority improvement item list, retrieve and combine from the conclusion suggestion corpus to generate hierarchical improvement suggestions;

[0121] In this embodiment, step S64 aims to provide prioritized improvement suggestions based on the severity and importance of the diagnostic results. The priority improvement item list clearly defines the relative importance of each improvement. One implementation method is to pre-define suggestion entries in a conclusion suggestion corpus and label each entry with tags such as its domain, impact level, and implementation difficulty. Based on the specific items in the priority improvement item list, matching suggestions are retrieved from the corpus and combined and presented according to priority order. Another implementation method utilizes semantic matching and knowledge graph technology to match the problem descriptions in the priority improvement item list with the solutions in the conclusion suggestion corpus, and automatically generates targeted, tiered improvement suggestions based on preset rules or models.

[0122] S65: Based on the preset report structure logic, integrate, logically verify, and render narrative text paragraphs, data visualization charts, and hierarchical improvement suggestions to generate an enterprise operation diagnostic report.

[0123] In this embodiment, step S65 aims to ensure that all independently generated report elements can be organized according to a predetermined structure and logic, and ultimately presented as a complete report. The report structure logic defines the report's chapter order, content layout, and interrelationships. One implementation method is to use a modular report generation framework, which defines various areas of the report, such as summary, problem analysis, data support, and improvement suggestions. This framework allows generated narrative text paragraphs, data visualization charts, and hierarchical improvement suggestions to be filled into their respective areas, while also performing logical checks, such as verifying data consistency and whether suggestions match the problems. Another implementation method is to utilize an automated typesetting and rendering engine to format the integrated content according to preset report templates and style guidelines, generating enterprise operation diagnostic reports in PDF, HTML, or other formats, ensuring the report's professionalism and readability.

[0124] Specifically, the solution in this application integrates a report intelligence agent to analyze the comprehensive diagnostic conclusions obtained in the preceding steps. First, the comprehensive diagnostic conclusions are analyzed to accurately extract key entities, quantitative values, and the logical relationships between them. Then, based on the extracted information, semantic adaptation and filling are performed with a pre-selected text narrative template to generate accurate and fluent narrative text paragraphs, ensuring the professionalism and readability of the report content. Simultaneously, for the core problem attributions and priority improvement item lists in the comprehensive diagnostic conclusions, the most effective chart types for expressing this information are matched and instantiated from the data visualization chart component library, and the relevant quantitative values ​​and logical relationships are accurately bound to the chart data source, enabling complex diagnostic data to be presented in an intuitive visual form. Furthermore, based on the ranking of the priority improvement item list, improvement suggestions with clear priorities are retrieved and combined from the conclusion suggestion corpus to generate clear action guidelines for enterprises. Finally, all generated narrative text paragraphs, data visualization charts, and hierarchical improvement suggestions are integrated, logically verified, and rendered based on the preset report structure logic, ensuring the overall structure of the report is rigorous, the content is coherent, and the logic is self-consistent, thereby generating a high-quality enterprise operation diagnostic report.

[0125] Through the above technical solutions, this application can achieve refined analysis of comprehensive diagnostic conclusions, ensuring a high degree of consistency between the report content and the diagnostic results. Specifically, through semantic adaptation and filling, the generated narrative text paragraphs can be made fluent and natural, avoiding a stiff splicing feel and improving the readability of the report. At the same time, data visualization charts are automatically matched and instantiated according to the characteristics of the diagnostic conclusions, enabling complex data information to be presented in an intuitive and easy-to-understand way, enhancing the persuasiveness of the report. In addition, graded improvement suggestions are generated according to priority, providing enterprises with clear action guidelines. Finally, through the integration, logical verification, and rendering of all report elements, the overall structure of the enterprise operation diagnostic report is ensured to be rigorous, the content coherent, and the logic self-consistent, improving the efficiency and quality of report generation and providing more effective support for enterprise decision-making.

[0126] It should be noted that the report integration agent in this embodiment is a dedicated software entity used to automatically transform structured comprehensive diagnostic conclusions into readable, reliable, and rigorously structured enterprise operation diagnostic reports. Its core function is to understand the semantics of the diagnostic conclusions and intelligently associate and synthesize them with various elements in the report knowledge base, rather than simply filling in templates.

[0127] For example, the report integration agent may include: a conclusion parsing engine, a narrative generation engine, a chart rendering engine, a suggestion synthesis engine, a report assembly engine, and a report knowledge base upon which they depend. The conclusion parsing engine can be configured to perform structured parsing of the input comprehensive diagnostic conclusions, identifying and extracting core problem attributions, overall health scores, key business entities included in the priority improvement item list, quantitative indicator values, and causal or comparative logical relationships between entities. The narrative generation engine interacts with a text narrative template library in the report knowledge base. Based on the output of the conclusion parsing engine, it can match the most suitable text template for different conclusion types and severity levels, and adapt and fill the parsed entities, values, and logical relationships into the corresponding slots of the templates using natural language generation technology, forming coherent and professional narrative text paragraphs. The chart rendering engine interacts with a data visualization chart component library in the report knowledge base. Based on the data characteristics of the core problem attributions and the priority improvement item list, such as trend comparisons, component proportions, and item rankings, it can dynamically select and instantiate the most effective chart type, such as line charts, bar charts, and pie charts, and bind the relevant quantitative values ​​and logical relationships into the chart. The data source of the table generates visual chart objects with correct data mapping relationships; the suggestion synthesis engine interacts with the conclusion suggestion corpus in the report knowledge base. Based on the priority list of improvement items and the associated repair cost assessment value of each item, it retrieves corresponding measure fragments, implementation points, and early warning information from the corpus, and combines and refines them according to the priority and logical relationship of the improvement items to generate targeted and hierarchical improvement suggestion chapters; the report assembly engine is the final report synthesis and quality control module. Following a preset report logic structure, such as "Summary - Problem Analysis - Data Presentation - Improvement Suggestions," it integrates the outputs of the narrative generation engine, chart rendering engine, and suggestion synthesis engine. During this process, the report assembly engine performs cross-chapter logical consistency checks, such as ensuring consistency between the data mentioned in the text and the chart data, and calls document formatting tools for overall layout and rendering, outputting a final formatted enterprise operation diagnostic report; the report knowledge base is a structured storage system that can contain reusable text narrative templates, parameterized chart components, modular conclusion suggestion corpora, and report structure specifications, providing material and rule support for the above engines.

[0128] Meanwhile, other architectural approaches for integrating intelligent agents into reports can be modified and configured according to actual scenarios and needs, and will not be elaborated here.

[0129] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0130] In one embodiment, a multi-agent collaborative enterprise diagnosis and operation report generation system is provided, which corresponds one-to-one with the multi-agent collaborative enterprise diagnosis and operation report generation method described in the previous embodiment. The multi-agent collaborative enterprise diagnosis and operation report generation system includes:

[0131] The data acquisition module is used to respond to enterprise diagnostic requests and acquire multi-dimensional operational data of the target enterprise corresponding to the enterprise diagnostic request.

[0132] The intelligent agent invocation module is used to perform time-series unified processing of multi-dimensional operational data, generate analysis datasets, and invoke several corresponding professional analytical intelligent agents in parallel based on enterprise diagnostic requests.

[0133] The sub-result generation module is used to synchronously input the analysis dataset into the professional analysis agent to obtain several diagnostic sub-results, including quantitative indicators, problem localization, and qualitative evaluation.

[0134] The diagnostic conclusion generation module is used to integrate and resolve conflicts in the obtained diagnostic scores according to a preset integration logic to generate a comprehensive diagnostic conclusion, which includes core problem attribution, overall health score and priority improvement item list.

[0135] The template component matching module is used to match the corresponding text description template, data visualization chart component, and conclusion suggestion corpus from the pre-set report knowledge base based on the comprehensive diagnostic conclusion;

[0136] The report generation module is used to fill in and associate the comprehensive diagnostic conclusions with the selected text description templates, data visualization chart components, and conclusion suggestion corpora through the report integration intelligent agent, and generate an enterprise operation diagnostic report.

[0137] Specific limitations regarding the multi-agent collaborative enterprise diagnosis and operation report generation system can be found in the above-described limitations of the multi-agent collaborative enterprise diagnosis and operation report generation method, and will not be repeated here. Each module in the aforementioned multi-agent collaborative enterprise diagnosis and operation report generation system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0138] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a method for generating enterprise diagnostic and operational reports based on multi-agent collaboration.

[0139] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements a method for generating enterprise diagnostic and operational reports based on multi-agent collaboration.

[0140] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for generating enterprise diagnostic and operational reports based on multi-agent collaboration, characterized in that, Including the following steps: In response to enterprise diagnostic requests, obtain multi-dimensional operational data of the target enterprise corresponding to the enterprise diagnostic request; Perform time-series unified processing on multidimensional operational data to generate an analysis dataset, and invoke several corresponding professional analytical agents in parallel based on enterprise diagnostic requests; The analysis dataset is synchronously input into a professional analytical agent to obtain several diagnostic scores, including quantitative indicators, problem localization, and qualitative evaluation. The obtained diagnostic scores are fused and conflict-resolved according to the preset integration logic to generate a comprehensive diagnostic conclusion, which includes core problem attribution, overall health score and priority improvement item list. Based on the comprehensive diagnostic conclusions, the corresponding text description templates, data visualization chart components, and conclusion suggestion corpora are matched from the pre-set report knowledge base; By integrating the report with an intelligent agent, the comprehensive diagnostic conclusions are populated and correlated with the selected text narrative templates, data visualization chart components, and conclusion suggestion corpora to generate an enterprise operation diagnostic report.

2. The method for generating enterprise diagnostic and operational reports based on multi-agent collaboration according to claim 1, characterized in that: The step of obtaining multidimensional operational data of the target enterprise corresponding to the enterprise diagnostic request in response to the enterprise diagnostic request includes: Analyze enterprise diagnostic requests to determine the business area and time frame; Based on a defined business area and time frame, first operational data is collected from the business systems associated with the target enterprise through a preset data interface. The first operational data includes financial transaction data, production record data, and human resource data. Secondary operational data is collected from external data sources associated with the target company through a data collection terminal. This secondary operational data includes industry market data, publicly available data, and upstream and downstream supply chain data. Data cleaning and field standardization are performed on the first and second operational data to form multidimensional operational data.

3. The method for generating enterprise diagnostic and operational reports based on multi-agent collaboration according to claim 1, characterized in that: The steps of performing time-series unified processing on multidimensional operational data to generate an analysis dataset, and then invoking several corresponding professional analytical agents in parallel based on enterprise diagnostic requests, include: Identify the original time stamp and data frequency corresponding to each data item in multidimensional operational data; Data items with different original time stamps and data frequencies are converted to a preset standard diagnostic time granularity and aligned to a preset time series reference point to obtain the analysis dataset; The system parses enterprise diagnostic requests, identifies several target diagnostic dimensions, and matches and invokes the corresponding professional analytical agents from a predefined set of professional analytical agents based on these dimensions.

4. The method for generating enterprise diagnostic and operational reports based on multi-agent collaboration according to claim 3, characterized in that: The steps of parsing the enterprise diagnostic request, determining several target diagnostic dimensions, and matching and calling the corresponding professional analytical agents from a predefined set of professional analytical agents based on the determined target diagnostic dimensions include: Semantic recognition is performed on the enterprise diagnostic request to extract business domain keywords and diagnostic description keywords; The extracted business domain keywords are mapped and matched with the preset diagnostic knowledge base. Combined with the diagnostic description keywords, several target diagnostic dimensions and their invocation priorities are determined. Based on the determined target diagnostic dimensions and their invocation priorities, the corresponding professional analytical agents are invoked sequentially from the predefined set of professional analytical agents.

5. The method for generating enterprise diagnostic and operational reports based on multi-agent collaboration according to claim 1, characterized in that: The steps of fusing and resolving conflicts in the obtained diagnostic scores according to a preset integration logic to generate a comprehensive diagnostic conclusion, which includes core problem attribution, overall health score, and a list of priority improvement items, include: Based on the multi-dimensional evaluation system and related knowledge network associated with the pre-set integration logic, the obtained diagnostic scores are structured and semantically aligned to obtain several aligned scores. Based on the enterprise's diagnostic request, the corresponding historical diagnostic conclusions are obtained, and the quantitative indicators in the alignment results are mapped to a unified measurement space in the multidimensional evaluation system. The mapping results are then calibrated based on the feedback data of the historical diagnostic conclusions. By inferring the coupling relationship and influence strength among various diagnostic dimensions in the current diagnostic context through the associated knowledge network, dynamic contribution weights are assigned to the calibrated alignment score based on the coupling relationship and influence strength. Identify the points of disagreement in the qualitative evaluation and problem localization of each alignment result, and based on the causal relationship and dynamic contribution weight revealed by the associated knowledge network, perform evidence fusion and collaborative reasoning on the points of disagreement to generate a consistent judgment. Based on consistency judgment, an overall health score is generated through a pre-set synthesis function in a multi-dimensional evaluation system. The core problem attribution is located by tracing the key paths in the related knowledge network, and a priority improvement item list is generated by combining dynamic contribution weights with a pre-set repair cost assessment model.

6. The method for generating enterprise diagnostic and operational reports based on multi-agent collaboration according to claim 5, characterized in that: The steps of generating an overall health score based on consistency judgment using a pre-set composition function in a multi-dimensional evaluation system, locating the core problem attribution by tracing the critical path in the related knowledge network, and generating a priority improvement item list by comprehensively combining dynamic contribution weights with a pre-set repair cost assessment model include: Based on consistency judgment, quantitative indicators are extracted from each aligned score result after calibration and dynamic contribution weighting; The extracted quantitative indicators are input into the pre-set synthesis function of the multidimensional evaluation system for weighted aggregation, and the confidence of the weighted aggregation result is adjusted by combining dynamic contribution weights to generate an overall health score. Identify key issues in consistency judgment, traverse the related knowledge network starting from the key issues, and trace the lower-level related nodes that lead to the key issues; Based on the traced lower-level related nodes, the influence intensity of each path is identified, the path with the largest cumulative influence intensity is identified as the critical path, and the core problem is located and attributed based on the critical path. By combining the dynamic contribution weights, the cost assessment values ​​output by the repair cost assessment model, and the impact propagation range assessment values ​​of the critical path, a list of priority improvement items is generated.

7. The method for generating enterprise diagnostic and operational reports based on multi-agent collaboration according to claim 1, characterized in that: The steps involved in generating an enterprise operation diagnostic report by integrating the comprehensive diagnostic conclusions with the selected text description templates, data visualization chart components, and conclusion suggestion corpora through a report integration intelligent agent include: The comprehensive diagnostic conclusions are analyzed to extract key entities, quantitative values, and logical relationships. Based on the extracted key entities, quantified values, and logical relationships, semantic adaptation and filling are performed with the selected text narrative template to generate narrative text paragraphs. Based on the characteristics of the core problem attribution and the priority improvement item list, the corresponding chart type is matched and instantiated from the selected data visualization chart component library, and the relevant quantitative values ​​and logical relationships are bound to the chart data source to obtain the data visualization chart; Based on the order of the priority improvement item list, hierarchical improvement suggestions are generated by retrieving and combining them from the conclusion suggestion corpus; Based on the preset report structure logic, the system integrates, logically verifies, and renders narrative text paragraphs, data visualization charts, and tiered improvement suggestions to generate an enterprise operation diagnostic report.

8. A multi-agent collaborative enterprise diagnosis and operation report generation system, characterized in that, include: The data acquisition module is used to respond to enterprise diagnostic requests and acquire multi-dimensional operational data of the target enterprise corresponding to the enterprise diagnostic request. The intelligent agent invocation module is used to perform time-series unified processing of multi-dimensional operational data, generate analysis datasets, and invoke several corresponding professional analytical intelligent agents in parallel based on enterprise diagnostic requests. The sub-result generation module is used to synchronously input the analysis dataset into the professional analysis agent to obtain several diagnostic sub-results, including quantitative indicators, problem localization, and qualitative evaluation. The diagnostic conclusion generation module is used to integrate and resolve conflicts in the obtained diagnostic scores according to a preset integration logic to generate a comprehensive diagnostic conclusion, which includes core problem attribution, overall health score and priority improvement item list. The template component matching module is used to match the corresponding text description template, data visualization chart component, and conclusion suggestion corpus from the pre-set report knowledge base based on the comprehensive diagnostic conclusion; The report generation module is used to fill in and associate the comprehensive diagnostic conclusions with the selected text description templates, data visualization chart components, and conclusion suggestion corpora through the report integration intelligent agent, and generate an enterprise operation diagnostic report.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for generating enterprise diagnostic and operational reports based on multi-agent collaboration as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for generating enterprise diagnostic and operational reports based on multi-agent collaboration as described in any one of claims 1-7.

Images