A financial management system and method based on big data

Abstract

The present invention belongs to the field of financial management, and discloses a financial management system and method based on big data; the present invention collects multi-source financial data, financial data images and comprehensive financial information of the company to construct a multimodal financial management model, the client receives the multi-source financial data to be processed and the financial data images to be processed, and generates the multi-source financial vector features to be processed by multimodal fusion, and the constructed multimodal financial management model generates the required comprehensive financial information according to the multi-dimensional financial management features to be processed, and the financial related personnel need to assist in judgment according to the comprehensive financial information, so as to make corresponding early warning decisions. The present invention combines financial data from multiple sources and uses financial image processing technology to achieve the optimized collection of relevant financial data using artificial intelligence big data, reduce the workload and statistical error rate of financial personnel, provide more accurate and comprehensive financial information, and help enterprises better cope with various challenges.

Description

Technical Field

[0001] The present invention belongs to the technical field of financial management, and in particular relates to a financial management system and method based on big data. Background Art

[0002] Traditional financial management systems usually use simpler methods to process and store financial data, such as using spreadsheets or small databases. These methods may be sufficient when processing small amounts of data, but when processing large-scale, complex, and multi-source financial data, their efficiency and accuracy often cannot meet the needs of modern enterprises.

[0003] With the rapid development of big data technology, the application of big data in the field of financial management has gradually attracted attention. Big data technology can process massive amounts of data, provide more accurate and comprehensive financial information, and help companies make better decisions. However, the existing financial management system based on big data still has some challenges and limitations in data acquisition, storage, analysis and early warning.

[0004] Existing technologies obtain comprehensive financial data through the system, perform data detection, discover data anomalies and errors, improve data accuracy and reliability, or record correction parameters to help the system update and correct data in real time to improve data quality. By monitoring and analyzing key parameters in the data correction process, identifying possible problems and delays, and helping financial personnel to promptly discover and solve data correction problems, the accuracy and efficiency of data processing can be improved. All of the above are to avoid financial risks by constantly correcting erroneous data to obtain more accurate financial data, which is essentially still using manual processing methods.

[0005] In order to obtain a smarter big data financial data management method, how to combine financial data from multiple sources and process financial-related images to achieve optimized collection of relevant financial data using artificial intelligence big data, reduce the workload of financial personnel and statistical error rate, provide more accurate and comprehensive financial information, and help companies better respond to various challenges are the difficulties faced by existing technologies.

[0006] The present invention proposes a financial management system and method based on big data. The system constructs a multimodal financial management model by collecting the company's multi-source financial data, financial data images and comprehensive financial information, and receives the multi-source financial data to be processed and the financial data images to be processed, and generates the multi-source financial vector features to be processed by multimodal fusion. The constructed multimodal financial management model generates the required comprehensive financial information according to the multi-dimensional financial management features to be processed, and the financial related personnel need to assist in judgment based on the comprehensive financial information, so as to make corresponding early warning decisions. Summary of the invention

[0007] In order to solve the above technical problems, the present invention proposes a financial management system and method based on big data.

[0008] In a first aspect of the present invention, a financial management method based on big data is provided, characterized in that the method comprises the following steps:

[0009] P1. Acquire the company's first multi-source financial data, call the first financial data image taken by the image acquisition device, perform image processing on the first multi-source financial data to obtain first multi-source financial image data, and also acquire the first comprehensive financial information at this time;

[0010] P2. Process the first financial data image and the first multi-source financial image data to form a first multi-source financial vector feature corresponding to a multi-dimensional label vector;

[0011] P3. Constructing a multimodal financial management model using the first multi-source financial vector features and the first comprehensive financial information, wherein the multimodal financial management model adopts a neural network model;

[0012] P4. Receive the company's second multi-source financial data and second financial data image, process the second multi-source financial data and the second financial data image, generate second multi-source financial image data and obtain second multi-source financial vector features in combination with the second financial data image, the multimodal financial management model processes the second multi-source financial vector features to generate second comprehensive financial information, and financial-related personnel assist in judgment based on the second comprehensive financial information to provide more accurate and comprehensive financial information.

[0013] Furthermore, the first multi-source financial data or the second multi-source financial data includes document financial data in Word format, Excel format and PDF format.

[0014] Furthermore, the processing of the first financial data image and the first multi-source financial image data to form a first multi-source financial vector feature corresponding to a multi-dimensional label vector or the processing of the second financial data image and the second multi-source financial image data to form a second multi-source financial vector feature corresponding to a multi-dimensional label vector is performed by a similarity method, specifically using the Pearson correlation coefficient method to calculate the correlation between the first financial data image and the font part of the first multi-source financial image data or the correlation between the second financial data image and the font part of the second multi-source financial image data.

[0015] Furthermore, taking C as a threshold, if C is greater than 0, the multi-source font image is identified by using a text recognition technology, and the first multi-source financial vector feature or the second multi-source financial vector feature corresponding to the project is listed according to the project required by the financial needs. The first multi-source financial vector feature or the second multi-source financial vector feature vector is expressed as ,in , , 、....、 is the item field value of h items, , , 、....、 For Project , , 、....、 Corresponding multi-source financial vector features.

[0016] Furthermore, the text recognition technology uses Baidu's OCR to perform text recognition in images.

[0017] Furthermore, the neural network model adopts an improved CNN neural network model, and its improved activation function calculation formula is as follows:

[0018]

[0019] is the activation function value, X is the input multi-source financial vector feature, h is the number of items required by the finance, is the value of the lth correlation, and m is the number of correlations C calculated using the similarity method.

[0020] A financial management system based on big data is also provided, the system includes a financial document acquisition module, a document conversion image module, a multi-source financial vector feature processing module, a multi-modal financial management model module, and a comprehensive financial acquisition module, characterized in that:

[0021] The financial document acquisition module is used to acquire the company's first multi-source financial data and the company's second multi-source financial data; the first multi-source financial data or the second multi-source financial data includes document financial data in Word format, Excel format and PDF format;

[0022] The document conversion image module is used to convert the first multi-source financial data or the second multi-source financial data into an image representation, and call the first financial data image or the second financial data image taken by the image acquisition device;

[0023] The multi-source financial vector feature processing module is used to process the first financial data image and the first multi-source financial image data to form a first multi-source financial vector feature corresponding to a multi-dimensional label vector; and is also used to process the second financial data image and the second multi-source financial image data to form a second multi-source financial vector feature corresponding to a multi-dimensional label vector;

[0024] The multimodal financial management model module is used to construct a multimodal financial management model using the first multi-source financial vector features and the first comprehensive financial information, wherein the multimodal financial management model adopts a neural network model;

[0025] The comprehensive financial acquisition module receives the company's second multi-source financial data and second financial data image, and transmits the second multi-source financial data and the second financial data image to the multi-source financial vector feature processing module, processes the second multi-source financial data and the second financial data image, generates second multi-source financial image data and combines the second financial data image to obtain second multi-source financial vector features, calls the multimodal financial management model to process the second multi-source financial vector features to generate second comprehensive financial information, and financial-related personnel assist in judgment based on the second comprehensive financial information to provide more accurate and comprehensive financial information.

[0026] Furthermore, the processing of the first financial data image and the first multi-source financial image data to form a first multi-source financial vector feature corresponding to a multi-dimensional label vector or the processing of the second financial data image and the second multi-source financial image data to form a second multi-source financial vector feature corresponding to a multi-dimensional label vector is performed by a similarity method, specifically using the Pearson correlation coefficient method to calculate the correlation between the first financial data image and the font part of the first multi-source financial image data or the correlation between the second financial data image and the font part of the second multi-source financial image data.

[0027] Furthermore, taking C as a threshold, if C is greater than 0, the multi-source font image is identified by using a text recognition technology, and the first multi-source financial vector feature or the second multi-source financial vector feature corresponding to the project is listed according to the project required by the financial needs. The first multi-source financial vector feature or the second multi-source financial vector feature vector is expressed as ,in , , 、....、 is the item field value of h items, , , 、....、 For Project , , 、....、 Corresponding multi-source financial vector features.

[0028] Furthermore, the text recognition technology uses Baidu's OCR to perform text recognition in images;

[0029] The neural network model adopts an improved CNN neural network model, and its improved activation function calculation formula is as follows:

[0030]

[0031] is the activation function value, X is the input multi-source financial vector feature, h is the number of items required by the finance, is the value of the lth correlation, and m is the number of correlations C calculated using the similarity method.

[0032] The present invention proposes a financial management system and method based on big data. The system constructs a multimodal financial management model by collecting multi-source financial data, financial data images and comprehensive financial information of the company, normalizes the multi-source financial data, so that comprehensive financial information can be obtained by processing only image data, receives the multi-source financial data to be processed and the financial data image to be processed, and generates the multi-source financial vector features to be processed by multimodal fusion. Specifically, the image similarity method is used to collect the fonts with high similarity in the picture and convert them into the value of the financial project based on this, so as to reduce the intensity of manual screening work, and the constructed multimodal financial management model generates the required comprehensive financial information according to the multi-dimensional financial management features to be processed. The model adopts the CNN model with improved activation function, so that the model has stronger generalization ability for the application scenario of financial management, and the financial related personnel can automatically generate comprehensive financial information according to the collection and processing according to the needs of financial information to assist in judgment, so as to make corresponding early warning decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a flow chart of a financial management method based on big data of the present invention;

[0034] Figure 2 It is a schematic diagram of the structure of a financial management system based on big data of the present invention;

[0035] Figure 3 It is a schematic diagram of the project field value and the corresponding vector eigenvalue in the present invention;

[0036] Figure 4 It is a schematic diagram of the principle of the Pearson correlation coefficient in the present invention. DETAILED DESCRIPTION

[0037] The invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0038] The first embodiment of the present invention:

[0039] The present invention provides a financial management system and method based on big data.

[0040] In a first aspect of the present invention, Figure 1 As shown, the present invention provides a financial management method based on big data, characterized in that the method comprises the following steps:

[0041] P1. Acquire the company's first multi-source financial data, call the first financial data image taken by the image acquisition device, perform image processing on the first multi-source financial data to obtain first multi-source financial image data, and also acquire the first comprehensive financial information at this time;

[0042] P2. Process the first financial data image and the first multi-source financial image data to form a first multi-source financial vector feature corresponding to a multi-dimensional label vector;

[0043] P3. Constructing a multimodal financial management model using the first multi-source financial vector features and the first comprehensive financial information, wherein the multimodal financial management model adopts a neural network model;

[0044] P4. Receive the company's second multi-source financial data and second financial data image, process the second multi-source financial data and the second financial data image, generate second multi-source financial image data and obtain second multi-source financial vector features in combination with the second financial data image, the multimodal financial management model processes the second multi-source financial vector features to generate second comprehensive financial information, and financial-related personnel assist in making judgments based on the second comprehensive financial information to provide more accurate and comprehensive financial information.

[0045] Furthermore, the first multi-source financial data or the second multi-source financial data includes document financial data in Word format, Excel format and PDF format.

[0046] Furthermore, the processing of the first financial data image and the first multi-source financial image data to form a first multi-source financial vector feature corresponding to a multi-dimensional label vector or the processing of the second financial data image and the second multi-source financial image data to form a second multi-source financial vector feature corresponding to a multi-dimensional label vector is performed by a similarity method, specifically using the Pearson correlation coefficient method to calculate the correlation between the first financial data image and the font part of the first multi-source financial image data or the correlation between the second financial data image and the font part of the second multi-source financial image data.

[0047] Specifically, the Pearson correlation coefficient method is used to calculate the correlation C between the first financial data image and the first multi-source financial image data font part or the second financial data image and the second multi-source financial image data font part:

[0048]

[0049] In the formula, each pixel is composed of three features of R, G, and B to form a feature vector, n pixels constitute the financial data image and the multi-source font image with the smallest number of pixels occupied by the font part of the multi-source financial image data, are the characteristic means of the vector composed of the three eigenvalues ​​of R, G, and B in the i-th pixel, and It is the feature mean composed of the feature values ​​R, G, and B of n pixels.

[0050] Furthermore, taking C as a threshold, if C is greater than 0, the multi-source font image is identified by using a text recognition technology, and the first multi-source financial vector feature or the second multi-source financial vector feature corresponding to the project is listed according to the project required by the financial needs. The first multi-source financial vector feature or the second multi-source financial vector feature vector is expressed as ,in , , 、....、 is the item field value of h items, , , 、....、 For Project , , 、....、 Corresponding multi-source financial vector features.

[0051] Furthermore, the text recognition technology uses Baidu's OCR to perform text recognition in images.

[0052] Furthermore, the neural network model adopts an improved CNN neural network model, and its improved activation function calculation formula is as follows:

[0053]

[0054] is the activation function value, X is the input multi-source financial vector feature, h is the number of items required by the finance, is the value of the lth correlation, and m is the number of correlations C calculated using the similarity method.

[0055] like Figure 2As shown, the present invention also provides a financial management system based on big data, the system includes a financial document acquisition module, a document conversion image module, a multi-source financial vector feature processing module, a multi-modal financial management model module, and a comprehensive financial acquisition module, characterized in that:

[0056] The financial document acquisition module is used to acquire the company's first multi-source financial data and the company's second multi-source financial data; the first multi-source financial data or the second multi-source financial data includes document financial data in Word format, Excel format and PDF format;

[0057] The document-to-image module is used to convert the first multi-source financial data or the second multi-source financial data into an image representation, and call the first financial data image or the second financial data image taken by the image acquisition device;

[0058] The multi-source financial vector feature processing module is used to process the first financial data image and the first multi-source financial image data to form a first multi-source financial vector feature corresponding to a multi-dimensional label vector; and is also used to process the second financial data image and the second multi-source financial image data to form a second multi-source financial vector feature corresponding to a multi-dimensional label vector;

[0059] The multimodal financial management model module is used to construct a multimodal financial management model using the first multi-source financial vector features and the first comprehensive financial information, wherein the multimodal financial management model adopts a neural network model;

[0060] The comprehensive financial acquisition module receives the company's second multi-source financial data and second financial data image, and transmits the second multi-source financial data and the second financial data image to the multi-source financial vector feature processing module, processes the second multi-source financial data and the second financial data image, generates second multi-source financial image data and combines the second financial data image to obtain second multi-source financial vector features, calls the multimodal financial management model to process the second multi-source financial vector features to generate second comprehensive financial information, and financial-related personnel assist in judgment based on the second comprehensive financial information to provide more accurate and comprehensive financial information.

[0061] like Figure 3 As shown, the present invention lists some project field values ​​and corresponding vector eigenvalues;

[0062] Furthermore, the correlation C between the first financial data image and the first multi-source financial image data font part or the second financial data image and the second multi-source financial image data font part is calculated by using the Pearson correlation coefficient method:

[0063]

[0064] In the formula, each pixel is composed of three features of R, G, and B to form a feature vector, n pixels constitute the financial data image and the multi-source font image with the smallest number of pixels occupied by the font part of the multi-source financial image data, are the characteristic means of the vector composed of the three eigenvalues ​​of R, G, and B in the i-th pixel, and It is the feature mean composed of the feature values ​​R, G, and B of n pixels.

[0065] like Figure 4 As shown, the present invention provides a schematic diagram of the principle of the Pearson correlation coefficient.

[0066] Furthermore, taking C as a threshold, if C is greater than 0, the multi-source font image is identified by using a text recognition technology, and the first multi-source financial vector feature or the second multi-source financial vector feature corresponding to the project is listed according to the project required by the financial needs. The first multi-source financial vector feature or the second multi-source financial vector feature vector is expressed as ,in , , 、....、 is the item field value of h items, , , 、....、 For Project , , 、....、 Corresponding multi-source financial vector features.

[0067] Furthermore, the text recognition technology uses Baidu's OCR to perform text recognition in images;

[0068] The neural network model adopts an improved CNN neural network model, and its improved activation function calculation formula is as follows:

[0069]

[0070] is the activation function value, X is the input multi-source financial vector feature, h is the number of items required by the finance, is the value of the lth correlation, and m is the number of correlations C calculated using the similarity method.

[0071] The comprehensive financial information output in this embodiment is the financial items selected by the financial-related candidates and their corresponding values.

[0072] The present invention proposes a financial management system and method based on big data. The system constructs a multimodal financial management model by collecting the company's multi-source financial data, financial data images and comprehensive financial information. By normalizing the multi-source financial data, the system only needs to process the image data to obtain the comprehensive financial information. By receiving the multi-source financial data to be processed and the financial data image to be processed, the multi-modal fusion generates the multi-source financial vector features to be processed. The constructed multimodal financial management model generates the required comprehensive financial information according to the multi-dimensional financial management features to be processed. The model adopts a CNN model with an improved activation function, so that the model has a stronger generalization ability for the application scenario of financial management. Financial-related personnel can automatically generate comprehensive financial information according to the collection and processing according to the needs of financial information to assist in judgment, so as to make corresponding early warning decisions.

[0073] A combination of multiple embodiments of the present invention can achieve all of the above effects, but it is not required that each embodiment of the present invention achieve all of the above advantages and effects, because each embodiment of the present invention can constitute an independent technical solution and make one or more contributions to the prior art.

[0074] The present invention does not particularly specify the structure of some modules, which shall be subject to the contents recorded in the prior art. The prior art mentioned in the above background technology section and the specific embodiment section of the present invention can be used as part of the present invention to understand the meaning of some technical features or parameters. The scope of protection of the present invention shall be subject to the contents actually recorded in the claims.

Claims

1. A financial management method based on big data, characterized in that: The method comprises the following steps: P1. Acquire the company's first multi-source financial data, call the first financial data image taken by the image acquisition device, perform image processing on the first multi-source financial data to obtain first multi-source financial image data, and also acquire the first comprehensive financial information at this time; P2. Process the first financial data image and the first multi-source financial image data to form a first multi-source financial vector feature corresponding to a multi-dimensional label vector; P3. Constructing a multimodal financial management model using the first multi-source financial vector features and the first comprehensive financial information, wherein the multimodal financial management model adopts a neural network model; P4. Receive the company's second multi-source financial data and second financial data image, process the second multi-source financial data and the second financial data image, generate second multi-source financial image data and obtain second multi-source financial vector features in combination with the second financial data image, process the second multi-source financial vector features by the multimodal financial management model to generate second comprehensive financial information, and financial-related personnel assist in making judgments based on the second comprehensive financial information to provide more accurate and comprehensive financial information; Processing the financial data image and the multi-source financial image data to form a multi-source financial vector feature corresponding to a multi-dimensional label vector specifically includes: using a Pearson correlation coefficient method to calculate the correlation C of the financial data image and the multi-source font image composed of the font part of the multi-source financial image data; Take C as the threshold. If C is greater than 0, the multi-source font image is identified by using the text recognition technology, and the first multi-source financial vector feature or the second multi-source financial vector feature corresponding to the project is listed according to the project required by the financial needs. The first multi-source financial vector feature or the second multi-source financial vector feature vector is expressed as ,in , , 、....、 is the item field value of h items, , , 、....、 For Project , , 、....、 Corresponding multi-source financial vector features; The neural network model adopts an improved CNN neural network model, and its improved activation function calculation formula is as follows: is the activation function value, X is the input multi-source financial vector feature, h is the number of items required by the finance, is the value of the lth correlation, and m is the number of correlations C calculated using the similarity method.

2. A financial management method based on big data as claimed in claim 1, characterized in that: The first multi-source financial data or the second multi-source financial data includes document financial data in Word format, Excel format and PDF format.

3. A financial management method based on big data as claimed in claim 1, characterized in that: The text recognition technology uses Baidu's OCR to perform text recognition on images.

4. A financial management system based on big data, which implements the method according to claim 1, and comprises a financial document acquisition module, a document conversion image module, a multi-source financial vector feature processing module, a multi-modal financial management model module, and a comprehensive financial acquisition module, characterized in that: The financial document acquisition module is used to acquire the company's first multi-source financial data and the company's second multi-source financial data; the first multi-source financial data or the second multi-source financial data includes document financial data in Word format, Excel format and PDF format; The document conversion image module is used to convert the first multi-source financial data or the second multi-source financial data into an image representation, and call the first financial data image or the second financial data image taken by the image acquisition device; The multi-source financial vector feature processing module is used to process the first financial data image and the first multi-source financial image data to form a first multi-source financial vector feature corresponding to a multi-dimensional label vector; and is also used to process the second financial data image and the second multi-source financial image data to form a second multi-source financial vector feature corresponding to a multi-dimensional label vector; The multimodal financial management model module is used to construct a multimodal financial management model using the first multi-source financial vector features and the first comprehensive financial information, wherein the multimodal financial management model adopts a neural network model; The comprehensive financial acquisition module receives the second multi-source financial data and the second financial data image of the company, transmits the second multi-source financial data and the second financial data image to the multi-source financial vector feature processing module, processes the second multi-source financial data and the second financial data image, generates second multi-source financial image data and obtains second multi-source financial vector features in combination with the second financial data image, calls the multimodal financial management model to process the second multi-source financial vector features to generate second comprehensive financial information, and financial-related personnel assist in making judgments based on the second comprehensive financial information to provide more accurate and comprehensive financial information; Processing the financial data image and the multi-source financial image data to form a multi-source financial vector feature corresponding to a multi-dimensional label vector specifically includes: using a Pearson correlation coefficient method to calculate the correlation of the financial data image and the multi-source font image composed of the font part of the multi-source financial image data; Take C as the threshold. If C is greater than 0, the multi-source font image is identified by using the text recognition technology, and the first multi-source financial vector feature or the second multi-source financial vector feature corresponding to the project is listed according to the project required by the financial needs. The first multi-source financial vector feature or the second multi-source financial vector feature vector is expressed as ,in , , 、....、 is the item field value of h items, , , 、....、 For Project , , 、....、 Corresponding multi-source financial vector features; The text recognition technology uses Baidu's OCR to perform text recognition on images; The neural network model adopts an improved CNN neural network model, and its improved activation function calculation formula is as follows: is the activation function value, X is the input multi-source financial vector feature, h is the number of items required by the finance, is the value of the lth correlation, and m is the number of correlations C calculated using the similarity method.

Images