Method and system for generating credit enhancement data report based on large model Agent agent

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a method and a system for generating a credit-increased data report based on a large model Agent agent. Financial data are automatically processed through the credit-increased large model Agent agent, the report generation efficiency can be remarkably improved, the manual operation time and potential errors are reduced, the Agent is combined with a Visual BERT model, and the efficiency of report generation is improved. According to the technical scheme, financial data input from different modalities can be processed, the accuracy and customization of the generated augmented credit report can be improved, and the technical problems that in the prior art, tedious manual operation is needed in the data collection, arrangement, analysis and report writing stages of the generated augmented credit data report, time is consumed, and human errors are likely to be generated are solved.

Description

Technical Field

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

[0002] In the field of traditional financial management practice, generating an enhanced credit data report is a crucial task of inestimable significance. For the internal management of an enterprise, this report is an important basis for accurately understanding the enterprise's operating conditions and reasonably planning strategic decisions. It can help them clearly understand the enterprise's capital flow, asset status, and potential risks, so as to formulate more targeted and forward-looking development strategies. For external investors, the enhanced credit data report is a key reference for them to evaluate the investment value of the enterprise and judge investment risks. A true and reliable report can enhance their confidence in the enterprise and attract more capital investment. For financial institutions, it is an important voucher for determining whether to provide loans to enterprises, determining the loan amount and interest rate, directly affecting the enterprise's financing ability and financing cost. However, this process currently faces many challenges. Specifically, the entire process highly relies on cumbersome manual operations, covering multiple links from data collection, data collation, data analysis to report writing.

[0003] However, in the prior art, in the stages of data collection, collation, analysis, and report writing for generating an enhanced credit data report, all require cumbersome manual operations, which are both time-consuming and prone to human errors. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for generating an enhanced credit data report based on a large model Agent intelligent agent, aiming to solve the technical problem that in the prior art, in the stages of data collection, collation, analysis, and report writing for generating an enhanced credit data report, all require cumbersome manual operations, which are both time-consuming and prone to human errors.

[0005] To achieve the above purpose, a system for generating an enhanced credit data report based on a large model Agent intelligent agent according to the present invention includes a data access module, a preprocessing module, an intelligent analysis module, a feature mining module, a report generation module, and an interaction optimization module. The preprocessing module is connected to the data access module, the intelligent analysis module is connected to the preprocessing module, the feature mining module is connected to the intelligent analysis module, the report generation module is connected to the feature mining module, and the interaction optimization module is connected to the report generation module.

[0006] The present invention also provides a method for generating an enhanced credit data report based on a large model Agent intelligent agent, which is applied to the system for generating an enhanced credit data report based on a large model Agent intelligent agent as described above.

[0007] It includes the following steps:

[0008] Step 1: First, the data access module and the preprocessing module obtain the financial data of the enterprise in different formats based on the credit enhancement large model Agent intelligent agent, and clean and standardize the data;

[0009] Step 2: Subsequently, the intelligent analysis module and the feature mining module use the VisualBERT model to process the data containing visual and text information, and identify the key numerical values in the balance sheet, income statement, and cash flow statement;

[0010] Step 3: The identified key data is used as input and passed into the credit enhancement large model Agent intelligent agent for fine-tuning and pre-training;

[0011] Step 4: The report generation module uses the trained credit enhancement large model Agent intelligent agent to analyze the input data, and combines with the general requirements put forward by the user to generate the final credit enhancement data report;

[0012] Step 5: At the same time, the interaction optimization module continues to receive the user's requirements and prompt words, and the credit enhancement large model Agent intelligent agent is used to generate a more customized and accurate evaluation report.

[0013] Among them, in Step 1, when obtaining the financial data of the enterprise in different formats:

[0014] For chart data, the credit enhancement large model Agent intelligent agent uses image processing technology to extract the key visual elements and layout structures in the image;

[0015] For text data, the credit enhancement large model Agent intelligent agent performs word segmentation, standardization, and stop word removal on the text data.

[0016] Among them, in Step 2, the credit enhancement large model Agent intelligent agent inputs the preprocessed image feature vectors and text token sequences into the VisualBERT model. VisualBERT processes the input multi-modal data through multiple layers of Transformer, and uses the self-attention mechanism to capture the complex relationships between image regions and text elements at different levels, aligns the elements in the input text with the regions in the input image, and learns the joint representation between the image and the text;

[0017] Moreover, the credit enhancement large model Agent intelligent agent extracts the key financial indicators and identifies the key information of potential risk points according to the specific requirements for the financial data by the VisualBERT model.

[0018] Among them, in step three, the identified key data is used as input and fed into the credit enhancement large model Agent intelligent agent for fine-tuning and pre-training. At the same time, the credit enhancement large model Agent intelligent agent executes the task of pre-training the large model on financial data.

[0019] Among them, in step four, the report generation module uses the trained credit enhancement large model Agent intelligent agent to generate content, automatically generates text format or charts, and provides financial analysis results.

[0020] Among them, in step five, the interaction optimization module utilizes the environmental perception feedback principle of the credit enhancement large model Agent intelligent agent. The user continues to input the desired credit enhancement report style and content title on the large model assistant. The credit enhancement large model assistant can quickly respond to the user's needs and make real-time modifications and supplements to the report style and content according to the user's feedback content.

[0021] A method and system for generating a credit enhancement data report based on a large model Agent intelligent agent. By automatically processing financial data through the credit enhancement large model Agent intelligent agent, the present invention can significantly improve the efficiency of report generation, reduce the time of manual operations and potential errors. By using Agent in combination with the VisualBERT model, it can process financial data inputs from different modalities, increasing the accuracy and customization of the generated credit enhancement report. In this way, it solves the technical problem in the prior art that in the stages of data collection, collation, analysis, and report writing for generating a credit enhancement data report, all require cumbersome manual operations, which are both time-consuming and prone to human errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] 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 use in 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, without creative efforts, other drawings can be obtained based on these drawings.

[0023] Figure 1 is a schematic block diagram of a system for generating a credit enhancement data report based on a large model Agent intelligent agent of the present invention.

[0024] Figure 2 is a flowchart of a method for generating a credit enhancement data report based on a large model Agent intelligent agent of the present invention.

[0025] Figure 3 is a flowchart of the VisualBERT model of the present invention for financial data analysis and processing.

[0026] 101 - Data access module, 102 - Pre - processing module, 103 - Intelligent analysis module, 104 - Feature mining module, 105 - Report generation module, 106 - Interaction optimization module. Detailed implementation manners

[0027] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the drawings. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present invention, and should not be construed as a limitation to the present invention.

[0028] Please refer to Figure 1 , Figure 1 which is the principle block diagram of the credit - enhanced data report generation system based on the large - model Agent intelligent agent of the present invention.

[0029] The present invention provides a credit - enhanced data report generation system based on the large - model Agent intelligent agent, including a data access module 101, a pre - processing module 102, an intelligent analysis module 103, a feature mining module 104, a report generation module 105, and an interaction optimization module 106. The pre - processing module 102 is connected to the data access module 101, the intelligent analysis module 103 is connected to the pre - processing module 102, the feature mining module 104 is connected to the intelligent analysis module 103, the report generation module 105 is connected to the feature mining module 104, and the interaction optimization module 106 is connected to the report generation module 105.

[0030] Please refer to Figure 2 and Figure 3 , Figure 2 which is the flowchart of the credit - enhanced data report generation method based on the large - model Agent intelligent agent of the present invention. Figure 3 which is the flowchart of the VisualBERT model of the present invention for financial data analysis and processing.

[0031] The present invention also provides a credit - enhanced data report generation method based on the large - model Agent intelligent agent, which is applied to the credit - enhanced data report generation system based on the large - model Agent intelligent agent as described above.

[0032] The method includes the following steps:

[0033] S1. First, the data access module 101 and the pre - processing module 102 obtain financial data in different formats of an enterprise based on the credit - enhanced large - model Agent intelligent agent, and perform data cleaning and standardization processing on the data;

[0034] For this specific implementation manner, when obtaining financial data in different formats of an enterprise:

[0035] For chart data, the Agent uses image processing technology to extract key visual elements and layout structures in the image, uses the pre-trained image encoder ResNet to extract image features, and regards these features as unordered input tokens;

[0036] For text data, the Agent performs word segmentation and converts it into an embedding representation that the BERT model can understand. Integrate the image features and text embeddings to form multimodal input data.

[0037] S2. Subsequently, the intelligent analysis module 103 and the feature mining module 104 use the VisualBERT model to process data containing visual and text information, and identify the key numerical values in the balance sheet, income statement, and cash flow statement;

[0038] For this specific embodiment, the credit enhancement large model Agent inputs the preprocessed image feature vectors and text token sequences into the VisualBERT model. VisualBERT processes the input multimodal data through multiple layers of Transformers, and uses the self-attention mechanism to capture the complex relationships between image regions and text elements at different levels, aligns the elements in the input text with the regions in the input image, and learns the joint representation between the image and the text;

[0039] Moreover, the credit enhancement large model Agent extracts key financial indicators and identifies key information on potential risk points according to the specific requirements for financial data by the VisualBERT model.

[0040] S3. Input the identified key data into the credit enhancement large model Agent for fine-tuning and pre-training;

[0041] For this specific embodiment, input the identified key data into the credit enhancement large model Agent for fine-tuning and pre-training. At the same time, the credit enhancement large model Agent performs the task of pre-training the large model on financial data.

[0042] S4. The report generation module 105 uses the trained credit enhancement large model Agent to analyze the input data, and combines the general requirements put forward by the user to generate the final credit enhancement data report;

[0043] For this specific embodiment, the report generation module 105 uses the trained credit enhancement large model Agent to generate content, and automatically generates text format or charts, and provides financial analysis results.

[0044] S5. Meanwhile, the interaction optimization module 106 continues to receive the user's requirements and prompt words, and the credit enhancement large model Agent intelligent body is used to generate a more customized and accurate evaluation report.

[0045] For this specific embodiment, the interaction optimization module 106 utilizes the environmental perception feedback principle of the credit enhancement large model Agent intelligent body. The user continues to input the desired credit enhancement report style and content title on the large model assistant. The credit enhancement large model assistant can quickly respond to the user's requirements and make real-time modifications and supplements to the report style and content according to the user's feedback content.

[0046] In the present invention, the VisualBERT model performs financial data analysis and processing as follows:

[0047] S201. Take the preprocessed image feature vectors and text token sequences as inputs and feed them into the VisualBERT model. Among them, the image feature vectors are extracted using the pre-trained image encoder ResNet. These features are regarded as unordered input tokens. The text data is tokenized and converted into an embedding representation that the BERT model can understand. Integrate the image features and text embeddings to form multi-modal input data;

[0048] S202. VisualBERT processes the input multi-modal data through multiple layers of Transformer. The model introduces a visual embedding F to represent an image. Each feature f\in F of the object detection model corresponds to a bounding box region of this image. This visual embedding F consists of three parts: (1) the visual feature f_o of the bounding box region obtained through CNN calculation; (2) the segment embedding f_s, indicating that this is an image embedding; (3) the position embedding f_p, expressing the spatial position of each region feature in the image;

[0049] S203. VisualBERT uses the self-attention mechanism of Transformer to align the elements of the input text and input image regions. The self-attention mechanism allows the model to internally align words and image regions. Calculate the dot product between the elements in the input sequence, evaluate the similarity or correlation degree between them, and perform weighted summation on the input sequence through the correlation to generate a new representation;

[0050] S204. The VisualBERT model takes the combination of the input vector and the context vector as the output. The output contains the information of the entire input sequence. Extract the key numerical values in the balance sheet, income statement, and cash flow statement, and use these key numerical values as inputs to be fine-tuned in the large model.

[0051] The large model is fine-tuned using LoRA, and the parameters of the original VisualBERT model remain fixed and are not directly updated. Only matrices A and B are trained, and they are both low-rank matrices. The product of A and B simulates the changes in the original parameters. During the inference stage, the product of BA is added to the original parameters. The original parameters W0 of the VisualBERT model remain unchanged, while the update amount ▲W is simulated by the product of B and A, that is, W = W0 + BA.

[0052] S205. The credit enhancement large model Agent intelligent agent executes the task of pre-training the large model on financial data, enabling the model to adapt to the specific language and visual patterns in the financial field.

[0053] Using the credit enhancement data report generation method and system based on the large model Agent intelligent agent of the present invention, the present invention can significantly improve the efficiency of report generation, reduce the time of manual operations and potential errors by automatically processing financial data through the credit enhancement large model Agent intelligent agent. By using the Agent combined with the VisualBERT model, it can process financial data inputs from different modalities, increasing the accuracy and customization of the generated credit enhancement reports. In this way, it solves the technical problem in the prior art that in the stages of data collection, collation, analysis, and report writing for generating credit enhancement data reports, all require cumbersome manual operations, which are both time-consuming and prone to human errors.

[0054] The above-disclosed is only a preferred embodiment of the present invention. Of course, it cannot be used to limit the scope of the rights of the present invention. Those of ordinary skill in the art can understand the whole or part of the process of implementing the above embodiment, and the equivalent changes made according to the claims of the present invention still fall within the scope covered by the invention.

Claims

1. A credit enhancement data report generation system based on a large model Agent intelligent agent, characterized in that it includes a data access module, a preprocessing module, an intelligent analysis module, a feature mining module, a report generation module, and an interaction optimization module. The preprocessing module is connected to the data access module, the intelligent analysis module is connected to the preprocessing module, the feature mining module is connected to the intelligent analysis module, the report generation module is connected to the feature mining module, and the interaction optimization module is connected to the report generation module.

2. A credit enhancement data report generation method based on a large model Agent intelligent agent, applied to the credit enhancement data report generation system based on a large model Agent intelligent agent as claimed in claim 1, characterized in that it includes the following steps: Step 1: First, the data access module and the preprocessing module obtain the financial data of the enterprise in different formats based on the credit enhancement large model Agent intelligent agent, and perform data cleaning and standardization processing. Step 2: Subsequently, the intelligent analysis module and the feature mining module use the VisualBERT model to process the data containing visual and text information, and identify the key numerical values in the balance sheet, income statement, and cash flow statement. Step 3: The identified key data is used as input and fed into the credit enhancement large model Agent intelligent agent for fine-tuning and pre-training. Step 4: The report generation module uses the trained credit enhancement large model Agent intelligent agent to analyze the input data, and combines with the general requirements put forward by the user to generate the final credit enhancement data report. Step 5: At the same time, the interaction optimization module continues to receive the user's requirements and prompt words, and the credit enhancement large model Agent intelligent agent is used to generate a more customized and accurate evaluation report.

3. The credit enhancement data report generation method based on a large model Agent intelligent agent as claimed in claim 2, characterized in that in Step 1, when obtaining the financial data of the enterprise in different formats: For chart data, the credit enhancement large model Agent intelligent agent uses image processing technology to extract the key visual elements and layout structure in the image. For text data, the credit enhancement large model Agent intelligent agent performs word segmentation, standardization, and stop word removal processing on the text data.

4. The credit enhancement data report generation method based on a large model Agent intelligent agent as claimed in claim 3, characterized in that in Step 2, the credit enhancement large model Agent intelligent agent inputs the preprocessed image feature vectors and text token sequences into the VisualBERT model. VisualBERT processes the input multi-modal data through multiple layers of Transformer, and uses the self-attention mechanism to capture the complex relationships between image regions and text elements at different levels, aligns the elements in the input text with the regions in the input image, and learns the joint representation between the image and the text. Moreover, according to the specific requirements for the financial data, the credit enhancement large model Agent intelligent agent extracts the key financial indicators and identifies the key information of potential risk points through the VisualBERT model.

5. The method for generating an enhanced credit data report based on a large model Agent intelligent agent according to claim 4, wherein In step three, the identified key data is used as input and passed into the enhanced credit large model Agent intelligent agent for fine-tuning and pre-training. At the same time, the enhanced credit large model Agent intelligent agent executes the task of pre-training the large model on financial data.

6. The method for generating an enhanced credit data report based on a large model Agent intelligent agent according to claim 5, wherein In step four, the report generation module uses the trained enhanced credit large model Agent intelligent agent to generate content, automatically generates text format or charts, and provides financial analysis results.

7. The method for generating an enhanced credit data report based on a large model Agent intelligent agent according to claim 6, wherein In step five, the interaction optimization module utilizes the environmental perception feedback principle of the enhanced credit large model Agent intelligent agent. The user continues to input the desired enhanced credit report style and content title on the large model assistant. The enhanced credit large model assistant can quickly respond to the user's needs and make real-time modifications and supplements to the report style and content according to the user's feedback content.

Images