Financial analysis method and system based on augmented credit model Agent agent

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a financial analysis method and system based on a credit increasing large model Agent agent, mass financial data can be automatically processed through the credit increasing large model Agent, full-process intellectualization is achieved from data preprocessing, feature extraction to result analysis, user requirements can be automatically recognized and analyzed, and user experience is improved. Accurate suggestions are quickly provided through dynamic analysis, and the complexity and error rate of manual participation are reduced. The intelligent assistant for financial analysis also makes the analysis more targeted and reliable by using a knowledge graph in the financial field and combining industry standards, policies and regulations and enterprise operation data.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular, to a financial analysis method and system based on an Agent intelligent body of an enhanced credit large model. Background Art

[0002] In recent years, the automation and intelligence of enterprise financial analysis have become the trend of the development of the financial industry. With the rapid development of technology, the requirements of enterprises for financial management are no longer limited to traditional bookkeeping and report generation, but pay more attention to data analysis, prediction and decision-making support. In addition, with the integration of technologies such as big data and artificial intelligence, enterprises expect financial analysis to be able to process massive data in real time, provide instant, accurate and in-depth financial insights, so as to help management quickly respond to market changes.

[0003] However, the financial analysis in the existing technology relies on fixed templates and rules, and it is difficult to quickly adapt to the complex business data and diverse needs of enterprises. Summary of the Invention

[0004] The purpose of the present invention is to provide a financial analysis method and system based on an Agent intelligent body of an enhanced credit large model, aiming to solve the technical problem that the financial analysis in the existing technology relies on fixed templates and rules and is difficult to quickly adapt to the complex business data and diverse needs of enterprises.

[0005] To achieve the above purpose, a financial analysis method based on an Agent intelligent body of an enhanced credit large model adopted by the present invention includes the following steps:

[0006] Step 1: Receive enterprise business data and external information based on the Agent intelligent body of the enhanced credit large model, and preprocess the data;

[0007] Step 2: Input the preprocessed financial data into the enhanced credit large model for fine-tuning to optimize the performance of the enhanced credit large model in the field of financial analysis;

[0008] Step 3: Use LLaMA 3 to parse the requirements of the user's query, accurately parse the user's intention, and then call the Agent for analysis and reasoning to complete the analysis task. Combine the dynamic analysis method, process the data in real time through prompt words and generate a response, so as to realize the real-time generation of professional financial analysis suggestions that meet the user's requirements;

[0009] Step 4: Maintain the knowledge graph in the field of financial analysis based on the Agent intelligent body of the enhanced credit large model to optimize the response to the user's query.

[0010] Among them, when preprocessing the data collected by the Agent intelligent body of the enhanced credit large model:

[0011] In the data cleaning section, for missing values, interpolation method is used for filling, and for outliers, they are replaced by combining industry average or median;

[0012] In the feature engineering section, random forest regression is used to process financial data to predict key indicators, and then key indicators in financial data (such as asset - liability ratio, current ratio, net profit margin, year - on - year / quarter - on - quarter growth rate, etc.) are extracted;

[0013] In the standardization processing section, continuous variables in financial data are normalized to the range of [0, 1] to reduce the impact of feature scale on the model.

[0014] Among them, when the pre - processed financial data is input into the credit enhancement large model, the credit enhancement large model Agent converts the data processed by random forest into text format (the converted format should be the text format suitable for the input of the credit enhancement large model) and constructs a fine - tuning data set, loads the basic large model using HuggingFace Transformers, and allows the credit enhancement large model to learn specific tasks of enterprise financial analysis through multiple rounds of training.

[0015] Among them, in step three, the specific method for generating financial analysis suggestions is as follows:

[0016] Use the LLaMA 3 model to parse user queries in the dialogue interaction stage, accurately understand user intentions, and extract core analysis tasks and data requirements through intention recognition and context analysis;

[0017] Coordinate other analysis and reasoning Agents through the dialogue Agent, call the data processing module and the reasoning module according to the parsing results, and complete the task decomposition and collaboration of user queries;

[0018] Combine dynamic analysis methods with prompt engineering, call the LLaMA 3 model for professional financial analysis based on real - time generated prompts, and dynamically process user requirements;

[0019] Based on the generated financial analysis suggestions, support real - time multi - round interaction, quickly respond to user feedback, and dynamically adjust analysis tasks to ensure the professionalism and pertinence of the generated results.

[0020] Among them, in step four, when the credit enhancement large model Agent maintains the knowledge graph in the field of financial analysis:

[0021] Adopt the knowledge base maintenance module to establish a knowledge graph with entities and relationships as the core based on industry standards, policies and regulations, and financial models, and use the LLaMA 3 model to automatically extract structured information from text and generate graph nodes and edges;

[0022] The Agent validates and optimizes the newly added knowledge according to the data verification mechanism to ensure the timeliness and correctness of the data;

[0023] In multiple rounds of interaction, the credit-enhanced large model Agent uses the associated information in the knowledge graph to generate professional financial analysis suggestions and optimize the response ability to user queries.

[0024] The present invention also provides a financial analysis system based on the credit-enhanced large model Agent, including a data preprocessing module, an analysis and reasoning module, a dialogue interaction module, and a knowledge base maintenance module for executing the financial analysis method based on the credit-enhanced large model Agent as described above.

[0025] When the financial analysis method and system based on the credit-enhanced large model Agent of the present invention are specifically used, first, the credit-enhanced large model Agent receives enterprise operation data and external information and preprocesses the data; the preprocessed financial data is input into the credit-enhanced large model for fine-tuning to optimize the performance of the credit-enhanced large model in the field of financial analysis; LLaMA 3 is used to parse the requirements of the user query, accurately parse the user's intention, and then call the analysis and reasoning Agent to complete the analysis task. Combining with the dynamic analysis method, the data is processed in real time through prompt words and a response is generated to realize the real-time generation of professional financial analysis suggestions that meet the user's requirements; the knowledge graph in the field of financial analysis is maintained based on the credit-enhanced large model Agent to optimize the response to user queries.

[0026] Through the credit-enhanced large model Agent, the present invention can automatically process a large amount of financial data, realize full-process intelligence from data preprocessing, feature extraction to result analysis, and can automatically identify and parse user requirements, quickly provide accurate suggestions through dynamic analysis, and reduce the complexity and error rate of manual participation. The financial analysis intelligent assistant also uses the knowledge graph in the financial field, combines industry standards, policies and regulations, and enterprise operation data to make the analysis more targeted and reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0028] Figure 1 is a flowchart of the financial analysis method based on the credit-enhanced large model Agent of the present invention.

[0029] Figure 2It is a flowchart for the Agent intelligent agent of the present invention to analyze user requirements.

[0030] Figure 3 It is a schematic diagram of the financial analysis system based on the credit-enhanced large model Agent intelligent agent of the present invention.

[0031] 101 - Data preprocessing module, 102 - Analysis and reasoning module, 103 - Dialogue interaction module, 104 - Knowledge base maintenance module. Detailed implementation manner

[0032] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to explain the present invention, and should not be construed as a limitation to the present invention.

[0033] Please refer to Figure 1 and Figure 2 where Figure 1 is a flowchart of the financial analysis method based on the credit-enhanced large model Agent intelligent agent of the present invention, Figure 2 is a flowchart for the Agent intelligent agent of the present invention to analyze user requirements.

[0034] The present invention provides a financial analysis method based on a credit-enhanced large model Agent intelligent agent, including the following steps:

[0035] S101. The credit-enhanced large model Agent intelligent agent receives enterprise operation data and external information, and preprocesses the data;

[0036] For this specific implementation manner, when preprocessing the data collected by the credit-enhanced large model Agent intelligent agent:

[0037] In the data cleaning part, for missing values, they are filled with interpolation method, and for outliers, they are replaced by combining industry average or median;

[0038] In the feature engineering part, random forest regression is used to process financial data, predict key indicators, and then extract key indicators from the financial data;

[0039] In the standardization processing part, continuous variables in the financial data are normalized to the range of [0, 1] to reduce the influence of feature scale on the model.

[0040] S102. Input the preprocessed financial data into the credit-enhanced large model for fine-tuning to optimize the performance of the credit-enhanced large model in the field of financial analysis;

[0041] For this specific embodiment, when the preprocessed financial data is input into the credit enhancement large model, the Agent intelligent agent of the credit enhancement large model converts the data processed by the random forest into text format and constructs a fine-tuning dataset, loads the basic large model using HuggingFace Transformers, and enables the credit enhancement large model to learn specific tasks of enterprise financial analysis through multiple rounds of training.

[0042] S103. Use LLaMA 3 to parse the requirements of the user query, accurately parse the user's intention, then call the analysis and reasoning Agent to complete the analysis task, combine with the dynamic analysis method, process the data in real time through prompt words and generate a response, so as to realize the real-time generation of professional financial analysis suggestions that meet the user's requirements;

[0043] For this specific embodiment, use the LLaMA 3 model to parse the user query in the dialogue interaction stage, accurately understand the user's intention, and extract the core analysis tasks and data requirements through intention recognition and context analysis;

[0044] Coordinate other analysis and reasoning Agents through the dialogue Agent, call the data processing module and the reasoning module according to the parsing result, and complete the task decomposition and cooperation of the user query;

[0045] Combine the dynamic analysis method and prompt engineering, call the LLaMA 3 model for professional financial analysis on the basis of real-time generation of prompt words, and dynamically process the user's requirements;

[0046] Based on the generated financial analysis suggestions, support real-time multi-round interaction, quickly respond to the user's feedback, and dynamically adjust the analysis tasks to ensure the professionalism and pertinence of the generated results.

[0047] S104. Based on the Agent intelligent agent of the credit enhancement large model, maintain the knowledge graph in the field of financial analysis and optimize the response to the user query.

[0048] For this specific embodiment, adopt the knowledge base maintenance module 104 to establish a knowledge graph with entities and relationships as the core based on industry standards, policies and regulations, and financial models, and use the LLaMA3 model to automatically extract structured information from the text and generate graph nodes and edges;

[0049] The Agent intelligent agent validates and optimizes the new knowledge according to the data verification mechanism to ensure the timeliness and correctness of the data;

[0050] In multi-round interaction, the Agent of the credit enhancement large model uses the associated information of the knowledge graph to generate professional financial analysis suggestions and optimize the response ability to the user query.

[0051] In the present invention, the process by which the Agent intelligent agent parses the user's requirements is as follows:

[0052] S201. The dialogue Agent analyzes the user's query requirements, combines context understanding, accurately analyzes the user's intention, and extracts key task requirements;

[0053] In this embodiment, specifically, it includes: when the user inputs "Please analyze the profitability of the company in the past three years", the LLaMA 3 model will analyze the user's query requirements into "task type", "data requirements", and "time range". The dialogue interaction Agent dynamically generates a task request according to the analysis result, calls the inference Agent to complete the analysis task, and utilizes the dialogue state tracking function of the LLaMA3 model in task allocation to ensure the context coherence of each step of the task;

[0054] S202. The dialogue Agent coordinates tasks, dynamically generates task requests, and calls other analysis Agents to assist in completing specific analysis tasks;

[0055] In this embodiment, specifically, it includes: the dialogue Agent issues an instruction to the data processing Agent, enabling the data processing Agent to process the input data by calling a random forest model according to the parsed intention. After the data processing Agent completes the task, it returns the specific result to the dialogue Agent, such as returning the financial indicator net profit margin.

[0056] The dialogue Agent sends the indicator to the inference Agent and requests it to generate an analysis report such as the profitability of the company.

[0057] S203. The Agent uses a dynamic analysis method to generate real-time financial analysis suggestions that meet the user's needs through prompt engineering. Based on the real-time generated prompts, it calls the LLaMA 3 model for professional financial analysis and dynamically processes the user's needs;

[0058] In this embodiment, specifically, it includes: the user can input the prompt "You are a professional financial analysis assistant. Based on the data provided above, complete a more in-depth assessment of the company's profitability." LLaMA 3 understands the user's needs through the prompt and calls the financial knowledge in the knowledge base Agent to generate a response.

[0059] S204. The dialogue Agent returns the analysis result generated by LLaMA 3 to the user and adjusts the prompt or calls a new analysis task according to the user's further feedback to form a closed loop.

[0060] In this embodiment, specifically, it includes: after the financial assessment report is generated in the dialogue Agent instance, when the user wants to further understand whether the company's profit growth trend conforms to the industry average level, the user continues to input requirements to the dialogue Agent, and the dialogue Agent calls other required Agents again to regenerate the assessment report.

[0061] Please refer to Figure 3 , Figure 3 which is the schematic diagram of the financial analysis system based on the credit enhancement large model Agent of the present invention.

[0062] The present invention also provides a financial analysis system based on the credit enhancement large model Agent, including a data preprocessing module 101, an analysis and reasoning module 102, a dialogue interaction module 103, and a knowledge base maintenance module 104 for executing the financial analysis method based on the credit enhancement large model Agent as described above.

[0063] The data preprocessing module 101 receives and integrates business data such as internal financial statements and business data of the enterprise and external environment information. The user can upload financial data files on the financial analysis intelligent assistant or input information such as enterprise financial data in the form of questions.

[0064] The analysis and reasoning module 102 adopts a pre-trained large language model optimized for the financial field, supporting tasks such as financial data analysis, causal reasoning, and risk assessment.

[0065] The dialogue interaction module 103 is based on natural language processing technology, supporting the user to initiate queries, adjust analysis parameters, and obtain real-time answers and suggestions in the form of dialogue.

[0066] The knowledge base maintenance module 104 supports the construction and dynamic update of the knowledge graph in the field of financial analysis. The credit enhancement large model Agent can verify and optimize the new knowledge according to the data verification mechanism to ensure the timeliness and correctness of the data.

[0067] When using a financial analysis method and system based on the credit-enhanced large model Agent of the present invention, in specific use, first, the credit-enhanced large model Agent receives enterprise operation data and external information, and preprocesses the data; the preprocessed financial data is input into the credit-enhanced large model for fine-tuning to optimize the performance of the credit-enhanced large model in the field of financial analysis; LLaMA 3 is used to parse the requirements of the user's query, accurately parse the user's intention, and then call the Agent for analysis and reasoning to complete the analysis task. Combining with the dynamic analysis method, the data is processed in real time through prompts and responses are generated to realize the real-time generation of professional financial analysis suggestions that meet the user's requirements; based on the credit-enhanced large model Agent, a knowledge graph in the field of financial analysis is maintained to optimize the response to the user's query.

[0068] Through the credit-enhanced large model Agent, the present invention can automatically process a large amount of financial data, realizing full-process intelligence from data preprocessing, feature extraction to result analysis, and can automatically identify and parse user requirements, quickly provide accurate suggestions through dynamic analysis, and reduce the complexity and error rate of manual participation. The financial analysis intelligent assistant also uses the knowledge graph in the financial field, combines industry standards, policies and regulations, and enterprise operation data to make the analysis more targeted and reliable.

[0069] The above-disclosed is only a preferred embodiment of the present invention, and of course, the scope of the rights of the present invention cannot be limited thereby. Those of ordinary skill in the art can understand all or part of the processes of implementing the above embodiments, and the equivalent changes made according to the claims of the present invention still fall within the scope covered by the present invention.

Claims

1. A financial analysis method based on the credit-enhanced large model Agent intelligent agent, characterized in that it includes the following steps: Step 1: The credit-enhanced large model Agent intelligent agent receives enterprise operation data and external information, and preprocesses the data; Step 2: Input the preprocessed financial data into the credit-enhanced large model for fine-tuning to optimize the performance of the credit-enhanced large model in the field of financial analysis; Step 3: Use LLaMA 3 to parse the requirements of the user's query, accurately parse the user's intention, and then call the analysis and reasoning Agent to complete the analysis task. Combining with the dynamic analysis method, process the data in real time through prompt words and generate a response to achieve real-time generation of professional financial analysis suggestions that meet the user's requirements; Step 4: Based on the credit-enhanced large model Agent intelligent agent, maintain the knowledge graph in the field of financial analysis to optimize the response to the user's query.

2. The financial analysis method based on the credit-enhanced large model Agent intelligent agent according to claim 1, characterized in that when preprocessing the data collected by the credit-enhanced large model Agent intelligent agent: In the data cleaning part, for missing values, use the interpolation method to fill them, and for outliers, replace them in combination with the industry average or median; In the feature engineering part, use random forest regression to process financial data, predict key indicators, and then extract the key indicators in the financial data; In the standardization processing part, normalize the continuous variables in the financial data to the range of [0,1] to reduce the influence of the feature scale on the model.

3. The financial analysis method based on the credit-enhanced large model Agent intelligent agent according to claim 2, characterized in that when inputting the preprocessed financial data into the credit-enhanced large model, the credit-enhanced large model Agent intelligent agent converts the data processed by the random forest into text format and constructs a fine-tuning data set, uses HuggingFace Transformers to load the basic large model, and allows the credit-enhanced large model to learn specific tasks of enterprise financial analysis through multiple rounds of training.

4. The financial analysis method and system based on the credit-enhanced large model Agent intelligent agent according to claim 3, characterized in that in step 3, the specific way to generate financial analysis suggestions is as follows: Use the LLaMA 3 model to parse the user's query in the dialogue interaction stage, accurately understand the user's intention, and extract the core analysis tasks and data requirements through intention recognition and context analysis; Coordinate other analysis and reasoning Agents through the dialogue Agent, call the data processing module and the reasoning module according to the parsing result, and complete the task decomposition and cooperation of the user's query; Combine the dynamic analysis method and the prompt word engineering, call the LLaMA 3 model for professional financial analysis on the basis of real-time generation of prompt words, and dynamically process the user's needs; Based on the generated financial analysis suggestions, support real-time multi-round interaction, quickly respond to the user's feedback, and dynamically adjust the analysis tasks to ensure the professionalism and pertinence of the generated results.

5. The financial analysis method and system based on the credit-enhanced large model Agent intelligent agent according to claim 4, characterized in that In step four, when the credit enhancement large model Agent maintains the knowledge graph in the field of financial analysis: The knowledge base maintenance module is used to establish a knowledge graph centered on entities and relationships based on industry standards, policies and regulations, and financial models. The LLaMA 3 model is used to automatically extract structured information from the text and generate graph nodes and edges; The Agent intelligent agent verifies and optimizes the new knowledge according to the data verification mechanism to ensure the timeliness and correctness of the data; In multiple rounds of interaction, the credit enhancement large model Agent uses the associated information of the knowledge graph to generate professional financial analysis suggestions and optimize the response ability of user queries.

6. A financial analysis system based on a credit-enhanced large model Agent intelligent agent, characterized in that, It includes a data preprocessing module, an analysis and reasoning module, a dialogue interaction module, and a knowledge base maintenance module for executing the financial analysis method based on the credit enhancement large model Agent intelligent agent as described in claim 5.

Images