Financial data intelligent management method and system based on RPA robot

Abstract

This invention relates to the field of text analysis technology, specifically to a method and system for intelligent financial data management based on RPA robots. The method involves: obtaining Chinese character phrases and numeric prefix phrases based on the ASCII encoding features of text information; constructing phrase sequences based on the positional features of Chinese character phrases; constructing binary label sequences based on the distributional positional features of numeric prefix phrases; obtaining numeric distribution difference values ​​based on any two binary label sequences; obtaining text feature distances based on the position and length of identical phrases and numeric distribution difference values ​​in any two phrase sequences; and clustering based on all text feature distances to obtain different receipt datasets. This invention obtains the corresponding receipt dataset based on the text feature distance between the latest bank receipt and all receipt datasets; and uses the RPA robot corresponding to the corresponding receipt dataset to process the latest bank receipt data, improving the accuracy of financial data reading and synchronization.

Description

Technical Field

[0001] This invention relates to the field of text analysis technology, specifically to a method and system for intelligent management of financial data based on RPA robots. Background Technology

[0002] RPA uses robots as virtual labor to complete pre-set tasks. In practical applications, it achieves high-frequency, high-volume repetitive data processing through automatically executed scripts. Automatic synchronization of receipts is a typical example of high-frequency, repetitive data processing. Compared to manual processing, RPA robots can save time and costs while achieving higher accuracy.

[0003] The current process of using RPA robots to synchronize bank receipts into the financial system has certain shortcomings. RPA reads receipt data according to a fixed script. Receipt formats differ between different banks and transaction types, and even the templates for the same bank may vary in different regions. If two receipts differ in descriptions such as full name vs. name, business type vs. settlement method, remittance transaction date vs. date, or in text format, it will cause difficulties for the RPA robot in reading the data, and may even lead to errors, affecting the accuracy of the RPA robot in processing financial data. Summary of the Invention

[0004] To address the technical problem of RPA robots struggling to read data from receipts due to differences in receipt types, thus reducing the accuracy of financial data synchronization, the present invention aims to provide a method and system for intelligent financial data management based on RPA robots. The specific technical solution adopted is as follows: Retrieve text information from bank receipts; Chinese character phrases and number prefix phrases are obtained based on the ASCII encoding features of the text information; a phrase sequence of the bank receipt is constructed based on the positional features of the Chinese character phrases in the text information; a binary label sequence is constructed based on the distribution positional features of the number prefix phrases in the phrase sequence. The numerical distribution difference value is obtained based on the difference features of the binary label sequences corresponding to any two bank receipts; the text feature distance is obtained based on the positional difference features and length features of the same word groups in the word group sequences of any two bank receipts, as well as the numerical distribution difference value; clustering is performed based on the text feature distance between all bank receipts to obtain different receipt datasets; The latest bank receipt belongs to a specific dataset based on the text feature distance between the latest bank receipt and all receipt datasets; the latest bank receipt is then processed by the RPA robot corresponding to that dataset.

[0005] Furthermore, the step of obtaining Chinese character phrases and number prefix phrases based on the ASCII encoding features of the text information includes: The text information is split before and after the character with an ASCII code value of 58, and the characters with ASCII code values ​​lower than 100 are deleted to obtain different Chinese word groups in the text information; if the ASCII code values ​​of the two characters following any Chinese word group in the text information represent a colon and an Arabic numeral, then the any Chinese word group is a number prefix word group.

[0006] Furthermore, the step of constructing the phrase sequence of the bank receipt based on the positional features of the Chinese character phrases in the text information includes: The phrase sequence of the bank receipt is obtained by sorting all Chinese character phrases in the order in which they appear in the text information.

[0007] Further, the step of constructing a binary label sequence based on the distribution position features of the numerical prefix words in the word sequence includes: In the word sequence, the word groups with numeric prefixes are represented by a preset first constant, and the word groups with non-numeric prefixes are represented by a preset second constant, thereby obtaining the binary tag sequence corresponding to the word sequence.

[0008] Furthermore, the step of obtaining the numerical distribution difference value based on the difference characteristics of the binary label sequences corresponding to any two bank receipts includes: Count the number of identical tags at the same position in the binary tag sequence corresponding to any two bank receipts to obtain the number of identical tags; calculate the reciprocal of the number of identical tags and normalize it to obtain the numerical distribution difference value.

[0009] Furthermore, the step of obtaining the text feature distance based on the positional difference features and length features of the same word groups in the word sequence of any two bank receipts, and the numerical distribution difference value, includes: Calculate the sum of the text lengths of identical phrases in the phrase sequences of any two bank receipts to obtain the total length of identical text; calculate the reciprocal of the total length of identical text and normalize it to obtain the first text difference value; calculate the absolute value of the difference in the position of the identical phrases in the two phrase sequences to obtain the position difference value; calculate the average of the position difference values ​​of all identical phrases in any two bank receipts and normalize it to obtain the second text difference value; calculate the sum of the first text difference value, the second text difference value, and the numerical distribution difference value to obtain the text feature distance between any two bank receipts.

[0010] Furthermore, the step of clustering based on the text feature distance between all bank receipts to obtain different receipt datasets includes: Clustering is performed using the K-means clustering algorithm based on the text feature distance between all bank receipts to obtain different receipt datasets.

[0011] Furthermore, the step of obtaining the receipt dataset to which the latest bank receipt belongs based on the text feature distance between the latest bank receipt and all receipt datasets includes: Calculate the text feature distance between the cluster centers of the latest bank receipt and any receipt dataset to obtain the difference representation value; take the receipt dataset corresponding to the minimum value of the difference representation value as the receipt dataset to which the latest bank receipt belongs.

[0012] The present invention also proposes a financial data intelligent management system based on RPA robots, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement any of the steps of a financial data intelligent management method based on RPA robots.

[0013] The present invention has the following beneficial effects: In this invention, obtaining Chinese character phrases and number prefix phrases can determine the valid information in the text information used to identify the type of bank receipt. Since different types of bank receipts will result in differences in the positional distribution characteristics of Chinese character phrases and Arabic numerals, obtaining the phrase sequence of the bank receipt can determine the positional characteristics of different Chinese character phrases in the receipt, and then determine whether the receipt types are the same based on the positional characteristics of the same phrases in different bank receipts. Obtaining the binary label sequence can be used to characterize the positional characteristics of Arabic numerals in the bank receipt, and then determine whether the receipt types are the same based on the positional characteristics of Arabic numerals in different bank receipts. Obtaining the number distribution difference value can characterize the degree of difference in the positional distribution of Arabic numerals in two bank receipts, and can be used to analyze whether the two bank receipt types are the same. Obtaining text feature distance can characterize the text similarity between two bank receipts, thus distinguishing different types of bank receipts based on this distance. Obtaining receipt datasets allows for the categorization of bank receipts of the same type, enabling each dataset to be assigned a dedicated RPA robot for reading and synchronizing financial data, improving the accuracy of RPA robot reading. Obtaining the receipt dataset to which the latest bank receipt belongs determines its receipt type, allowing it to be assigned to the corresponding dedicated RPA robot for reading and synchronizing financial data, further improving the accuracy of financial data reading and synchronization. Attached Figure Description

[0014] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of a financial data intelligent management method based on RPA robots, provided as an embodiment of the present invention. Detailed Implementation

[0016] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a financial data intelligent management method and system based on RPA robots proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0018] The following description, in conjunction with the accompanying drawings, details a specific solution for an intelligent financial data management method and system based on RPA robots provided by this invention.

[0019] Please see Figure 1 The diagram illustrates a flowchart of a financial data intelligent management method based on an RPA robot, according to an embodiment of the present invention. The method includes the following steps: Step S1: Obtain the text information from the bank receipt.

[0020] When an RPA robot reads text information from bank receipts, it follows a fixed script. Receipt formats differ between banks and transaction types, and even within the same bank, templates may vary across regions. If two receipts differ in descriptions such as full name vs. name, transaction type vs. settlement method, or remittance date vs. date, or in text format, the RPA robot will struggle to read the data, potentially leading to errors. Therefore, to improve the accuracy of RPA robot processing bank receipt data, different RPA robots can be created for different receipt types, with each robot handling only one type of bank receipt. However, RPA robots cannot automatically distinguish between different types of bank receipts. When a bank receipt requires data synchronization, the type of the receipt needs to be determined to distribute it to the appropriate RPA robot, thus improving data reading accuracy.

[0021] When two bank receipts contain a large number of identical phrases, it means that the two bank receipts belong to the same type; therefore, the receipt type can be classified based on the phrase characteristics of a large number of historical bank receipts. First, the text information of the bank receipt is obtained. In this embodiment of the invention, the set of all ordered text information in the receipt PDF file is obtained through existing OCR optical character recognition technology. This technology can accurately identify the text information in the image and automatically segment non-connected text information in the image. The text information is extracted according to phrase format, with each phrase forming a phrase, in an extraction order from left to right and from top to bottom. For example: "Customer-specific receipt from a certain bank, currency: RMB, voucher number, payee, remittance remarks: payment for goods, transaction institution, serial number: 123456, amount (in words) RMB One Million Yuan Only," etc., text information in different positions on the receipt.

[0022] Step S2: Obtain Chinese character phrases and number prefix phrases based on the ASCII encoding features of the text information; construct the phrase sequence of the bank receipt based on the positional features of the Chinese character phrases in the text information; construct the binary label sequence based on the distribution positional features of the number prefix phrases in the phrase sequence.

[0023] After acquiring the text information of a large number of historical bank receipts of all types, the text information includes: bank name, payer, payee, bank name, date, and Arabic numerals for bank account number, amount, and transaction number. The Arabic numeral portion of the text information is irrelevant to the receipt type, only related to the transaction scenario. Therefore, during the receipt type analysis, it is necessary to remove the Arabic numeral portion of the text information to improve the accuracy of receipt classification. Furthermore, text information containing colons, such as "Remittance Bank: [Bank Name]" and "Remittance Remarks: Payment for Goods," has phrases before the colon belonging to fixed receipt phrases, while the text after the colon belongs to transaction scenario information. Therefore, it is necessary to split consecutive text messages to improve the accuracy of receipt classification. Thus, Chinese character phrases and numerical prefix phrases are obtained based on the ASCII encoding features of the text information.

[0024] Preferably, in this embodiment of the invention, the steps of obtaining Chinese character phrases and numeric prefix phrases include: splitting the text before and after a character with an ASCII code value of 58 in the text information. In the existing ASCII encoding table, the code value 58 represents a colon. Therefore, splitting the text before and after a character with a code value of 58 can obtain independent phrases. For example, in the remittance remarks: payment for goods, the text information is split to obtain two Chinese character phrases: remittance remarks and payment for goods. Characters with ASCII code values ​​lower than 100 in the text information are deleted to obtain different Chinese character phrases in the text information. Since the ASCII code values ​​of Chinese characters are relatively large, usually above 20,000, while the code values ​​of Arabic numerals are relatively small, below 100; and some symbols, such as parentheses, also have code values ​​below 100, in order to remove Arabic numerals and symbol information that do not belong to the receipt template, text information with code values ​​below 100 is deleted to obtain Chinese character phrases that can reflect the receipt type. In text information, if the ASCII code value of the two characters following any Chinese character phrase represents a colon and an Arabic numeral, then that Chinese character phrase is a numeric prefix phrase. Since the position of Arabic numerals varies in different receipt templates, obtaining the position of Arabic numerals in the text information can improve the accuracy of receipt classification. For example, in the serial number 123456, if the ASCII code value following the Chinese character phrase serial number represents a colon and an Arabic numeral, then the Chinese character phrase serial number is a numeric prefix phrase. By determining the position of the numeric prefix phrase in the text information, the position of the Arabic numerals can be determined.

[0025] Furthermore, after obtaining all the Chinese character phrases on the bank receipt, a phrase sequence for the bank receipt can be constructed based on the positional features of the Chinese character phrases in the text information. Specifically, this involves sorting all the Chinese character phrases according to their order of appearance in the text information to obtain the phrase sequence for the bank receipt. The Chinese character phrases in the phrase sequence follow the recognition order of OCR technology. The phrase sequence represents the order in which different phrases appear on the bank receipt, and the phrase sequence features of two bank receipts can be used to determine whether they belong to the same type of receipt. To further improve the accuracy of receipt classification, the position of Arabic numerals on the receipt can be analyzed; therefore, a binary label sequence can be constructed based on the distribution positional features of the numeral prefix phrases in the phrase sequence. Preferably, in this embodiment of the invention, the step of obtaining the binary tag sequence includes: in the word sequence, the word group with a numeric prefix is ​​represented by a preset first constant, and the word group with a non-numeric prefix is ​​represented by a preset second constant, thereby obtaining the binary tag sequence corresponding to the word sequence; in this embodiment of the invention, the preset first constant is constant 0, the preset second constant is constant 1, the binary tag sequence is composed of 0 and 1, and the position of the Arabic numeral in the bank receipt can be determined by determining the position of 0 in the binary tag sequence.

[0026] Step S3: Obtain the numerical distribution difference value based on the difference features of the binary label sequences corresponding to any two bank receipts; obtain the text feature distance based on the positional difference features and length features of the same word groups in the word group sequences of any two bank receipts, as well as the numerical distribution difference value; perform clustering based on the text feature distance between all bank receipts to obtain different receipt datasets.

[0027] After obtaining the phrase sequences and binary label sequences of different bank receipts, the bank receipts can be clustered based on the characteristics of the phrase sequences and the binary label sequences, grouping similar types into one category. Since the positional distribution of Arabic numerals differs among different bank receipts, the closer the positions of the Arabic numerals in two bank receipts, the more likely the two receipts are of the same type. Therefore, the difference value of the numeral distribution is obtained based on the difference characteristics of the binary label sequences corresponding to any two bank receipts. Preferably, in this embodiment of the invention, the step of obtaining the difference value of the numeral distribution includes: counting the number of identical labels at the same position in the binary label sequences corresponding to any two bank receipts, obtaining the number of identical labels; for example, if the binary label sequences corresponding to two bank receipts are 11001 and 11100, then the first, second, and fourth labels are identical, and the number of identical labels is 3. The more identical labels there are, the more likely the two bank receipts belong to the same type. Calculate the reciprocal of the number of identical labels and normalize it to obtain the numerical distribution difference value. The smaller the numerical distribution difference value, the more similar the Arabic numeral distribution positions of the two bank receipts are, and the more likely they belong to the same type. The larger the numerical distribution difference value, the greater the difference in the Arabic numeral distribution positions are, and the less likely they belong to the same type.

[0028] Furthermore, the more identical phrases appear in the phrase sequences of two bank receipts, the more likely the two receipts are of the same type. However, shorter phrases are more likely to be similar, such as "account number" and "bank name," which appear in most receipts and have less impact on classification. If a large number of longer phrases appear in the same sequence, it strongly suggests they are of the same type, such as "actual recipient's bank" and "payer's bank." However, different types of receipts have format differences. Even if the same phrases appear in two receipts, they may be located in different positions. One phrase might be third in the phrase sequence, while the same phrase might be eighth in another. Therefore, the length of the identical phrases alone is insufficient to determine whether two bank receipts belong to the same type; it is also necessary to analyze the positional differences of the identical phrases in different receipts. Since the phrase sequence in step S2 is constructed based on the positional order of Chinese phrases on the receipt, the closer the positions of identical phrases in the phrase sequence, the more likely the two receipts are of the same type, not only because the Chinese phrases are identical but also because their positions on the bank receipts are similar. Therefore, the text feature distance can be obtained based on the positional differences of identical phrases in the phrase sequences of any two bank receipts, as well as the length characteristics and numerical distribution differences of the identical phrases.

[0029] Preferably, in this embodiment of the invention, the step of obtaining text feature distance includes: calculating the sum of the text lengths of identical phrases in the phrase sequences of any two bank receipts to obtain the total length of identical text, where the text length is the number of Chinese characters in the phrase; the longer the total text length, the more identical phrases appear in the two receipts and the longer their lengths. The reciprocal of the total length of identical text is calculated and normalized to obtain a first text difference value; the smaller the first text difference value, the more identical phrases appear in the two receipts and the longer the lengths of the identical phrases, making them more likely to be receipts of the same type. The absolute value of the difference in the position of identical phrases in the two phrase sequences is calculated to obtain a position difference value; the position represents the location of the Chinese character phrase in the phrase sequence, and the closer the positions of identical phrases in the two phrase sequences, the smaller the position difference value. Calculate the average and normalize the positional differences of all identical phrases on any two bank receipts to obtain the second textual difference value. A smaller second textual difference value indicates that the distribution of identical phrases on the two receipts is closer, making them more likely to be the same type of receipt. Calculate the sum of the first textual difference value, the second textual difference value, and the numerical distribution difference value to obtain the textual feature distance between any two bank receipts. A smaller textual feature distance indicates that the two bank receipts have more identical phrases, closer distributions, and closer distributions of Arabic numerals, making them more likely to belong to the same type of receipt. A larger textual feature distance indicates that the two bank receipts are less likely to belong to the same type. The formula for obtaining the textual feature distance includes: In the formula, This represents the text feature distance between the R-th bank receipt and the T-th bank receipt. This indicates normalization, and N represents the number of identical phrases in the two bank receipts. This represents the length of the text containing the nth identical phrase. This represents the total length of the same text, and 'a' represents a preset minimum positive number, which is 0.01 in this embodiment of the invention. This is to avoid cases where the denominator is 0; when the denominator is not 0, it is not included in the calculation. Indicates the first text difference value. This indicates the position of the nth identical phrase in the phrase sequence within the Rth bank receipt. This indicates the position of the nth identical phrase in the phrase sequence within the Tth bank receipt. Indicates the difference in rank. This represents the second text difference value, where H represents the number of identical tags. This represents the difference in the distribution of numbers.

[0030] Furthermore, after obtaining the text feature distances between all bank receipts, clustering can be performed based on these distances to obtain different receipt datasets. Preferably, in this embodiment, clustering is performed using the K-means clustering algorithm based on the text feature distances between all bank receipts to obtain different receipt datasets. It should be noted that the K-means clustering algorithm is existing technology, using the text feature distance as the distance between different receipt samples in the clustering space; the specific clustering steps will not be elaborated further. The bank receipts in each receipt dataset are of the same type, thus a dedicated RPA robot can be built based on the receipt types in that dataset. This dedicated RPA robot processes the bank receipts in that dataset, improving the accuracy of financial data reading and synchronization.

[0031] Step S4: Obtain the receipt dataset to which the latest bank receipt belongs based on the text feature distance between the latest bank receipt and all receipt datasets; organize the latest bank receipt data using the RPA robot corresponding to the receipt dataset to which it belongs.

[0032] When a new bank receipt becomes available in the future, the dataset to which the new bank receipt belongs can be determined by the text feature distance between the new bank receipt and all receipt datasets. Preferably, in this embodiment of the invention, the step of obtaining the dataset to which the new bank receipt belongs includes: calculating the text feature distance between the cluster centers of the new bank receipt and any receipt dataset to obtain a difference representation value; the smaller the difference representation value, the more likely the new bank receipt belongs to that any receipt dataset. The receipt dataset corresponding to the minimum difference representation value is taken as the dataset to which the new bank receipt belongs. Then, the RPA robot corresponding to the dataset to which the new bank receipt belongs performs data processing on the new bank receipt, improving the accuracy of the RPA robot's reading and synchronization of the financial data of the new bank receipt.

[0033] In summary, this invention provides a method for intelligent financial data management based on RPA robots. It obtains Chinese character phrases and numeric prefix phrases based on the ASCII encoding features of text information; constructs phrase sequences based on the positional features of Chinese character phrases; constructs binary label sequences based on the distributional positional features of numeric prefix phrases; obtains numeric distribution difference values ​​based on any two binary label sequences; obtains text feature distances based on the position and length of identical phrases and numeric distribution difference values ​​in any two phrase sequences; and performs clustering based on all text feature distances to obtain different receipt datasets. This invention obtains the corresponding receipt dataset based on the text feature distance between the latest bank receipt and all receipt datasets; and uses the RPA robot corresponding to the corresponding receipt dataset to process the latest bank receipt data, improving the accuracy of financial data reading and synchronization.

[0034] This invention also proposes a financial data intelligent management system based on RPA robots, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement any of the steps of a financial data intelligent management method based on RPA robots.

[0035] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0036] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for intelligent management of financial data based on an RPA robot, characterized in that, The method includes the following steps: Retrieve text information from bank receipts; Chinese character phrases and number prefix phrases are obtained based on the ASCII encoding features of the text information; a phrase sequence of the bank receipt is constructed based on the positional features of the Chinese character phrases in the text information; a binary label sequence is constructed based on the distribution positional features of the number prefix phrases in the phrase sequence. The numerical distribution difference value is obtained based on the difference features of the binary label sequences corresponding to any two bank receipts; the text feature distance is obtained based on the positional difference features and length features of the same word groups in the word group sequences of any two bank receipts, as well as the numerical distribution difference value; clustering is performed based on the text feature distance between all bank receipts to obtain different receipt datasets; The latest bank receipt belongs to a specific dataset based on the text feature distance between the latest bank receipt and all receipt datasets; the latest bank receipt is then processed by the RPA robot corresponding to that dataset. 2.The financial data intelligent management method based on the RPA robot according to claim 1, wherein, The steps of obtaining Chinese character phrases and number prefix phrases based on the ASCII encoding features of the text information include: The text information is split before and after the character with an ASCII code value of 58, and the characters with ASCII code values ​​lower than 100 are deleted to obtain different Chinese word groups in the text information; if the ASCII code values ​​of the two characters following any Chinese word group in the text information represent a colon and an Arabic numeral, then the any Chinese word group is a number prefix word group.

3. The intelligent financial data management method based on RPA robots according to claim 1, characterized in that, The step of constructing the word sequence of the bank receipt based on the positional features of the Chinese word groups in the text information includes: The phrase sequence of the bank receipt is obtained by sorting all Chinese character phrases in the order in which they appear in the text information.

4. The intelligent financial data management method based on RPA robots according to claim 1, characterized in that, The step of constructing a binary label sequence based on the distribution position features of the digital prefix words in the word sequence includes: In the word sequence, the word groups with numeric prefixes are represented by a preset first constant, and the word groups with non-numeric prefixes are represented by a preset second constant, thereby obtaining the binary tag sequence corresponding to the word sequence.

5. The intelligent financial data management method based on RPA robots according to claim 1, characterized in that, The step of obtaining the numerical distribution difference value based on the difference characteristics of the binary label sequences corresponding to any two bank receipts includes: Count the number of identical tags at the same position in the binary tag sequence corresponding to any two bank receipts to obtain the number of identical tags; calculate the reciprocal of the number of identical tags and normalize it to obtain the numerical distribution difference value.

6. The intelligent financial data management method based on RPA robots according to claim 1, characterized in that, The step of obtaining the text feature distance based on the positional difference features and length features of the same word groups in the word sequence of any two bank receipts, and the numerical distribution difference value includes: Calculate the sum of the text lengths of identical phrases in the phrase sequences of any two bank receipts to obtain the total length of identical text; calculate the reciprocal of the total length of identical text and normalize it to obtain the first text difference value; calculate the absolute value of the difference in the position of the identical phrases in the two phrase sequences to obtain the position difference value; calculate the average of the position difference values ​​of all identical phrases in any two bank receipts and normalize it to obtain the second text difference value; calculate the sum of the first text difference value, the second text difference value, and the numerical distribution difference value to obtain the text feature distance between any two bank receipts.

7. The intelligent financial data management method based on RPA robots according to claim 1, characterized in that, The step of clustering based on the text feature distance between all bank receipts to obtain different receipt datasets includes: Clustering is performed using the K-means clustering algorithm based on the text feature distance between all bank receipts to obtain different receipt datasets.

8. The intelligent financial data management method based on RPA robots according to claim 1, characterized in that, The step of determining the dataset to which the latest bank receipt belongs based on the text feature distance between the latest bank receipt and all receipt datasets includes: Calculate the text feature distance between the cluster centers of the latest bank receipt and any receipt dataset to obtain the difference representation value; take the receipt dataset corresponding to the minimum value of the difference representation value as the receipt dataset to which the latest bank receipt belongs.

9. A financial data intelligent management system based on RPA robots, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the method as described in any one of claims 1-8.

Images