Industrial multi-dimensional data fusion system and method

Abstract

The invention discloses an industrial multi-dimensional data fusion system and method, and relates to the technical field of electronic bill multi-dimensional monitoring, and the method comprises the following steps: building an electronic bill service platform, building an enterprise file for an internal enterprise of the platform based on MySQL, collecting the electronic bill information of the internal enterprise entering the platform, including a draft bill, a sales bill and a purchase bill, and storing the collected information in the platform. Recording the data in a related enterprise file; on the basis of electronic bill data in enterprises, information extraction is performed on bills, and bill data under archives of the enterprises of the two parties are retrieved for the enterprises entering the platform. According to the industrial multi-dimensional data fusion system and method, by fusing multi-dimensional data such as enterprise archives, electronic bill data and business range portraits, suspicious bill judgment is carried out on electronic bills, the business range portraits are dynamically generated based on actual business data of enterprises, the accuracy of determination of suspicious bills is enhanced, and the efficiency is improved. The workload of manual bill checking is reduced, and the practicability and functionality are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-dimensional monitoring of electronic bills, and in particular to an industrial multi-dimensional data fusion system and method. Background Art

[0002] Electronic bills are electronic forms of traditional paper bills, mainly bills generated in e-commerce, online payment and other transaction activities. Electronic bills include but are not limited to electronic invoices, electronic receipts, and electronic checks. Although the emergence of electronic bills has improved transaction efficiency, it has also provided criminals with new means of committing crimes. Therefore, it is necessary to conduct a suspiciousness analysis on electronic bills to determine whether the bill is a suspicious bill; However, the existing bill service platform determines the suspiciousness of electronic bills by manually checking the electronic bills one by one, which is highly subjective and labor-intensive. It is difficult to integrate the multi-dimensional data of both companies to quantitatively evaluate the suspiciousness of electronic bills. It has the problem of low practicality, and it is impossible to classify suspicious electronic bills, and it is impossible to further determine the suspiciousness of electronic bills based on the actual business profiles of both companies. It has the problem of low functionality. Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0003] In view of the problems in the related technology, the present invention proposes an industrial multi-dimensional data fusion system and method to overcome the above-mentioned technical problems existing in the existing related technology.

[0004] To this end, the specific technical solution adopted by the present invention is as follows: A method for fusion of multi-dimensional industrial data, the method comprising the following steps: S1. Build an electronic invoice service platform, establish enterprise archives for enterprises within the platform based on MySQL, collect electronic invoice information of enterprises within the platform, including bills of exchange, sales invoices, and purchase invoices, and record them in relevant enterprise archives; S2. Extract information from bills based on the internal electronic bill data of the enterprise. For the enterprises settled in the platform, retrieve the bill data under the archives of both enterprises, establish business scope portraits of both enterprises, determine suspicious bills based on the overlap of business portraits, and mark suspicious bills; S3. For enterprises not registered in the platform, retrieve all bill data stored in the platform, generate a list of related enterprises, generate a business scope profile based on the bill data in the files of related enterprises, determine suspicious bills and mark suspicious bills based on the determination results; S4. When there are no affiliated companies, collect the business scope of the company, combine it with the current business scope portrait of the company, judge the suspicious bills, and mark the suspicious bills based on the judgment results.

[0005] As a preferred embodiment, S2 includes the following sub-steps: S21. Extracting bill information based on the internal electronic bill data of the enterprise archives. The specific steps are as follows: For bills of exchange in corporate archives, search for keywords containing "payee" and "drawer" based on OCR tools, and extract the company name after the keywords; For sales invoices in corporate archives, use OCR tools to search for keywords containing "buyer" and "seller" and extract the company name after the keywords; For purchase invoices in corporate archives, use OCR tools to search for keywords containing "seller" and "buyer" and extract the company name after the keywords; S22. Based on the company names of both parties, retrieve the electronic invoice data under the company archives of both parties, build a portrait of the business scope of both companies, and determine the suspiciousness of the current electronic invoice based on the overlap of the business portraits.

[0006] As a preferred embodiment, the S22 includes the following sub-steps: S221. Based on the names of both companies, search the company files of both companies in the platform, summarize the sales invoice data and purchase invoice data in the company files of both companies, extract the product name field based on the OCR tool, and The extracted product names are classified by similarity, and the algorithm formula is: ; in, is the similarity, Represents the similarity threshold, when ≥ When , the two product names are classified into one category. < When , it means that the two commodities are not of the same category; S222. Calculate the weight ratios of different commodity categories for the classified commodity categories in the sales invoices and purchase invoices, and establish the purchase business profile and sales business profile of Enterprise A and Enterprise B. The specific steps are as follows: For purchase invoices, for each commodity category , add up the amounts of all purchase invoices under the commodity category, and get the total amount of all purchase invoices under the commodity category , and add up the total amount of all purchase invoices to get the total purchase amount , calculate the purchase weight proportion of different commodity categories respectively, and the algorithm formula is: ; in, Representative product categories Procurement weight ratio under the current enterprise; For sales invoices, for each product category , add up the amounts of all sales invoices under the product category, and get the total amount of all sales invoices under the product category , and add up the total amount of all sales invoices to get the total sales amount , calculate the sales weight proportion of different commodity categories respectively, and the algorithm formula is: ; in, Representative product categories Based on the sales weight ratio of the current enterprise, establish product categories for Enterprise A and Enterprise B respectively. The steps of purchasing business portrait and sales business portrait are as follows: Company A: , ; Company B: , ; in, Represents the purchasing weight of enterprise A in commodity category 1, represents the sales weight of enterprise A in product category 1, represents the purchasing weight of enterprise B in commodity category 1, Represents the sales weight of enterprise B in commodity category 1; S223. Based on the type of the current electronic bill, the business overlap between the current enterprise A and enterprise B is calculated based on the cosine similarity method. The specific steps are as follows: When the electronic bill is a bill of exchange of enterprise A and the payee is enterprise B, or the electronic bill is a purchase invoice of enterprise A and the seller is enterprise B, the similarity between the purchase business profile of enterprise A and the sales business profile of enterprise B is calculated. The specific algorithm formula is: ; in, Represents the similarity of purchasing business portraits. When the purchase business of company A and the sales business of company B do not match in scope, there is a mismatched transaction behavior, that is, the current electronic bill is a definite suspicious bill. This means that the purchasing business of enterprise A and the sales business of enterprise B are matched and have a basis for business cooperation, that is, the current electronic bill is a normal bill; When the electronic bill is a bill of exchange of enterprise A and the drawer is enterprise B, or the electronic bill is a sales invoice of enterprise A and the purchaser is enterprise B, the similarity between the sales business profile of enterprise A and the procurement business profile of enterprise B is calculated. The specific algorithm formula is: ; in, Represents the similarity of sales business portraits. When the sales business of company A and the purchasing business of company B do not match in scope, there is a mismatched transaction behavior, that is, the current electronic bill is a definite suspicious bill. This means that the sales business of enterprise A and the purchasing business of enterprise B are matched and have a basis for business cooperation, that is, the current electronic bill is a normal bill; S224. For confirmed abnormal electronic invoices, the relevant enterprise files shall be marked in red, and the confirmed abnormal electronic invoices shall also be marked in red.

[0007] As a preferred embodiment, S3 comprises the following steps: S31. Analyze the electronic invoice data according to the current enterprise electronic invoice type, the steps are as follows: When the enterprise information in the electronic invoice contains non-platform internal enterprises, extract the name of the non-platform internal enterprise , and retrieve all the bill data stored in the platform at the same time, and generate a list of related companies based on the names of non-platform companies; S32. Determine the suspiciousness of the current electronic invoice in combination with the list of related companies.

[0008] As a preferred implementation, the S31 includes the following sub-steps: S311. Traverse the internal enterprise archives of the platform to obtain the data set of all enterprise bills stored in the platform , from the ticket data set Filter out the names of companies that are not on the platform Notes: ; in, express A specific electronic ticket in For non-platform internal enterprises A subset of electronic invoices, from Extraction and Companies that have business dealings with the platform , the steps are: ; in, Represents the collection of enterprise names that have settled in the platform. For A collection of companies that have business dealings with the platform; S312. For each enterprise settled in the platform , calculate its The specific steps are: ; in, Represents the co-occurrence frequency, that is, the Collection of bills middle Number of occurrences; ; in, Represents the total transaction amount, which is calculated in the and The sum of all transaction amounts in the bill is based on and Calculating association strength The algorithm formula is: ; in, Represents the weight coefficient, according to the strength of association Sort the internal affiliated enterprises of the platform that are not currently settled in the platform in descending order of strength, and select the top The companies that have settled in the platform serve as business profile comparison companies for current electronic invoice companies.

[0009] As a preferred embodiment, the S32 comprises the following steps: S321, based on the previous For each enterprise settled in the platform, the method of step S22 is adopted to obtain business portraits of the business portrait comparison enterprises respectively, and the business portraits of all the business portrait comparison enterprises are merged to obtain business portraits of the associated enterprises; S322. Calculate the similarity between the business profile of the current electronic invoice enterprise and the business profile of the associated enterprise based on the cosine similarity. When the similarity is less than or equal to 0, it means that there is a mismatch between the current enterprise and the non-platform settled enterprise in the transaction behavior. When the similarity is greater than 0, it means that there is a business match between the current enterprise and the non-platform settled enterprise. S323. When the similarity is less than or equal to 0, the current electronic bill in the file is marked in blue, indicating that the current electronic bill is a questionable electronic bill, and further manual verification is performed afterwards.

[0010] As a preferred embodiment, the S4 includes the following sub-steps: S41. When there are no affiliated enterprises outside the platform, the business scope of enterprises outside the platform is collected, and the cosine similarity is determined based on the actual business profile of the current enterprise. When the similarity value is less than or equal to 0, it means that there is no business intersection between the current enterprise and the enterprises outside the platform, and the current electronic invoice is a confirmed suspicious invoice, which is marked in red in the enterprise file. When the similarity value is greater than 0, it means that the business scope of the enterprises outside the platform overlaps with the actual business scope of the current enterprise, and the current electronic invoice is recorded as a suspicious electronic invoice, which is marked in blue in the enterprise file and awaits further manual verification.

[0011] An industrial multi-dimensional data fusion system, including an enterprise data collection module, a business scope generation module, a suspiciousness determination module, and a manual verification module; The enterprise data collection module, based on the constructed electronic invoice service platform, establishes enterprise archives for enterprises within the platform through MySQL, collects electronic invoice information of enterprises within the platform, including bills of exchange, sales invoices, and purchase invoices, and records them in relevant enterprise archives; The business scope generation module extracts bill data based on bill data in different enterprise archives through OCR tools. The extracted bill commodity names are classified by similarity to generate actual business portraits of different enterprises; The suspiciousness determination module retrieves the bill data under the archives of both companies for the companies settled in the platform, determines suspicious bills based on the overlap of business portraits, and marks the confirmed suspicious bills in red. For companies not settled in the platform, it retrieves all bill data stored in the platform, generates a list of related companies, generates a portrait of the related business scope based on the bill data in the archives of the related companies, determines suspicious bills and marks them in blue. When there are no related companies, it collects the business scope of the companies, combines the current business scope portrait of the companies, determines the suspicious bills, and marks the confirmed suspicious bills in red and the suspicious bills in blue based on the determination results. The manual verification module further conducts manual verification on the suspicious bills marked in blue in the enterprise archives to determine the suspicious degree of the electronic bills.

[0012] The beneficial effects of the present invention are: 1. The present invention identifies the electronic invoice data of enterprises within the platform, and based on the information of both parties of the electronic invoice, analyzes the electronic invoice data of the enterprises under the jurisdiction of the platform, obtains the actual business portraits of both enterprises based on the scope of the goods under the jurisdiction of the invoices, quantifies the similarity of the business portraits of both enterprises, and more scientifically judges the suspiciousness of the invoices, determines the suspiciousness of the electronic invoices and marks the suspicious invoices, thereby reducing the workload of manual verification and enhancing practicality and functionality; 2. The present invention avoids the subjectivity and limitations of traditional manual judgment of suspiciousness by constructing the actual business portrait of the enterprise and quantifying the similarity of the business portrait. In addition, by constructing the actual business portrait of the enterprise and analyzing the similarity, the present invention identifies the bills that are inconsistent with the real business of both enterprises, promptly discovers suspicious bills and automatically marks them, reduces the workload of manual verification, and improves work efficiency. 3. The present invention analyzes the internal correlation of non-platform enterprises, thereby summarizing the business profiles of related enterprises within the platform, and performs preliminary screening and marking of the suspiciousness of electronic bills by comparing the similarity between the current enterprise business profile and the summarized business profile, thereby reducing the workload of manual verification, enhancing functionality, and judging the suspiciousness of bills more comprehensively and scientifically; 4. The present invention conducts a business scope query on non-platform enterprises that have no internal affiliated enterprises on the platform, and makes a judgment based on the actual business profile of the current enterprise, thereby classifying and judging the suspiciousness of electronic bills in different situations based on the enterprise type of electronic bills within the platform, thereby enhancing the accuracy of judging suspicious bills; The multi-dimensional data fusion system and method for the entire industry, by integrating multi-dimensional data such as enterprise archives, electronic invoice data, and business scope portraits, determines whether electronic invoices are suspicious, and dynamically generates business scope portraits based on the actual business data of the enterprise, thereby enhancing the accuracy of the determination of conclusive suspicious invoices, reducing the workload of manual invoice verification, and enhancing practicality and functionality. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0014] Figure 1 is a flow chart of an industrial multi-dimensional data fusion method according to an embodiment of the present invention; Figure 2 It is a block diagram of an industrial multi-dimensional data fusion system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0015] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0016] According to an embodiment of the present invention, a system and method for fusion of industrial multi-dimensional data are provided.

[0017] The present invention is further described with reference to the accompanying drawings and specific embodiments: Example 1: Figure 1 As shown, according to an industrial multi-dimensional data fusion method according to an embodiment of the present invention, the method includes the following steps: S1. Build an electronic invoice service platform, establish enterprise archives for enterprises within the platform based on MySQL, collect electronic invoice information of enterprises within the platform, including bills of exchange, sales invoices, and purchase invoices, and record them in relevant enterprise archives; S2. Extract information from bills based on the internal electronic bill data of the enterprise. For the enterprises settled in the platform, retrieve the bill data under the archives of both enterprises, establish business scope portraits of both enterprises, determine suspicious bills based on the overlap of business portraits, and mark suspicious bills; S21. Extracting bill information based on the internal electronic bill data of the enterprise archives. The specific steps are as follows: For bills of exchange in corporate archives, search for keywords containing "payee" and "drawer" based on OCR tools, and extract the company name after the keywords; For sales invoices in corporate archives, use OCR tools to search for keywords containing "buyer" and "seller" and extract the company name after the keywords; For purchase invoices in corporate archives, use OCR tools to search for keywords containing "seller" and "buyer" and extract the company name after the keywords; It should be noted that OCR tools include Tesseract, Baidu OCR, Google Cloud Vision, etc., which are used to perform text recognition on electronic invoices stored in corporate archives to extract the name of the other company of the corresponding invoice, providing data support for subsequent identification of suspicious invoices.

[0018] S22. Based on the names of both companies, retrieve the electronic invoice data under the archives of both companies, build a profile of the business scope of both companies, and determine the suspiciousness of the current electronic invoice based on the overlap of the business profiles; S221. Based on the names of both companies, search the company files of both companies in the platform, summarize the sales invoice data and purchase invoice data in the company files of both companies, extract the product name field based on the OCR tool, and The extracted product names are classified by similarity, and the algorithm formula is: ; in, is the similarity, Represents the similarity threshold, when ≥ When , the two product names are classified into one category. < When , it means that the two commodities are not of the same category; It should be noted that through Similarity is used to fuzzily match products. The value is generally set to 0.7, and can also be adjusted according to actual needs. The larger the similarity threshold, the more categories there are, and the more accurate the subsequent business scope division is.

[0019] S222. Calculate the weight ratios of different commodity categories for the classified commodity categories in the sales invoices and purchase invoices, and establish the purchase business profile and sales business profile of Enterprise A and Enterprise B. The specific steps are as follows: For purchase invoices, for each commodity category , add up the amounts of all purchase invoices under the commodity category, and get the total amount of all purchase invoices under the commodity category , and add up the total amount of all purchase invoices to get the total purchase amount , calculate the purchase weight proportion of different commodity categories respectively, and the algorithm formula is: ; in, Representative product categories Procurement weight ratio under the current enterprise; For sales invoices, for each product category , add up the amounts of all sales invoices under the product category, and get the total amount of all sales invoices under the product category , and add up the total amount of all sales invoices to get the total sales amount , calculate the sales weight proportion of different commodity categories respectively, and the algorithm formula is: ; in, Representative product categories Based on the sales weight ratio of the current enterprise, establish product categories for Enterprise A and Enterprise B respectively. The steps of purchasing business portrait and sales business portrait are as follows: Company A: , ; Company B: , ; in, Represents the purchasing weight of enterprise A in commodity category 1, represents the sales weight of enterprise A in product category 1, represents the purchasing weight of enterprise B in commodity category 1, Represents the sales weight of enterprise B in commodity category 1; It should be noted that the commodity category is the category of commodities sold or purchased by the current enterprise after calculation based on fuzzy matching. When discussing the purchasing weight ratio and sales weight ratio related to the commodity category, the same letters are used to indicate that the two calculations are based on the same commodity classification system.

[0020] S223. Based on the type of the current electronic bill, the business overlap between the current enterprise A and enterprise B is calculated based on the cosine similarity method. The specific steps are as follows: When the electronic bill is a bill of exchange of enterprise A and the payee is enterprise B, or the electronic bill is a purchase invoice of enterprise A and the seller is enterprise B, the similarity between the purchase business profile of enterprise A and the sales business profile of enterprise B is calculated. The specific algorithm formula is: ; in, Represents the similarity of purchasing business portraits. When the purchase business of company A and the sales business of company B do not match in scope, there is a mismatched transaction behavior, that is, the current electronic bill is a definite suspicious bill. This means that the purchasing business of enterprise A and the sales business of enterprise B are matched and have a basis for business cooperation, that is, the current electronic bill is a normal bill; When the electronic bill is a bill of exchange of enterprise A and the drawer is enterprise B, or the electronic bill is a sales invoice of enterprise A and the purchaser is enterprise B, the similarity between the sales business profile of enterprise A and the procurement business profile of enterprise B is calculated. The specific algorithm formula is: ; in, Represents the similarity of sales business portraits. When the sales business of company A and the purchasing business of company B do not match in scope, there is a mismatched transaction behavior, that is, the current electronic bill is a definite suspicious bill. This means that the sales business of enterprise A and the purchasing business of enterprise B are matched and have a basis for business cooperation, that is, the current electronic bill is a normal bill; It should be noted that and The range is between -1 and 1. and The closer the value is to 1, the higher the business overlap between Enterprise A and Enterprise B. By quantifying the similarity between Enterprise A's procurement business and Enterprise B's sales business, and quantifying the similarity between Enterprise A's sales business and Enterprise B's procurement business, a basis is provided for determining the suspiciousness of different types of electronic invoices.

[0021] S224. For confirmed abnormal electronic invoices, the relevant enterprise files are marked in red, and the confirmed abnormal electronic invoices are also marked in red; Embodiment 2: S3, for enterprises not settled in the platform, search all bill data stored in the platform, generate a list of related enterprises, generate a business scope portrait based on the bill data in the files of related enterprises, determine suspicious bills and mark suspicious bills based on the determination results; S31. Analyze the electronic invoice data according to the current enterprise electronic invoice type, the steps are as follows: When the enterprise information in the electronic invoice contains non-platform internal enterprises, extract the name of the non-platform internal enterprise , and retrieve all the bill data stored in the platform at the same time, and generate a list of related companies based on the names of non-platform companies; S311. Traverse the internal enterprise archives of the platform to obtain the data set of all enterprise bills stored in the platform , from the ticket data set Filter out the names of companies that are not on the platform Notes: ; in, express A specific electronic ticket in For non-platform internal enterprises A subset of electronic invoices, from Extraction and Companies that have business dealings with the platform , the steps are: ; in, Represents the collection of enterprise names that have settled in the platform. For A collection of companies that have business dealings with the platform; S312. For each enterprise settled in the platform , calculate its The specific steps are: ; in, Represents the co-occurrence frequency, that is, the Collection of bills middle Number of occurrences; ; in, Represents the total transaction amount, which is calculated in the and The sum of all transaction amounts in the bill is based on and Calculating association strength The algorithm formula is: ; in, Represents the weight coefficient, according to the strength of association Sort the internal affiliated enterprises of the platform that are not currently settled in the platform in descending order of strength, and select the top The companies that have settled in the platform serve as business profile comparison companies for current electronic invoice companies.

[0022] It should be noted that Co-occurrence frequency reflects and The frequency of co-occurrence in the bills, Represents the total transaction amount to reflect and The weight coefficient is used to balance the relative importance of co-occurrence frequency and total transaction amount in the calculation of association strength. The weight coefficient can be determined by analyzing historical data to determine the degree of influence of frequency and amount on association strength, or by consulting experts in related fields to set it. If the total transaction amount is considered to be more important than the co-occurrence frequency, a larger weight coefficient value can be selected. If the co-occurrence frequency is considered to be more important, a smaller weight coefficient value can be selected. The value is generally set to 5, and can also be adjusted according to actual needs. The smaller the value, the smaller the range of suspiciousness judgment.

[0023] S32. Determine the suspiciousness of the current electronic bill in combination with the list of related enterprises; S321, based on the previous For each enterprise settled in the platform, the method of step S22 is adopted to obtain business portraits of the business portrait comparison enterprises respectively, and the business portraits of all the business portrait comparison enterprises are merged to obtain business portraits of the associated enterprises; S322. Calculate the similarity between the business profile of the current electronic invoice enterprise and the business profile of the associated enterprise based on the cosine similarity. When the similarity is less than or equal to 0, it means that there is a mismatch between the current enterprise and the non-platform settled enterprise in the transaction behavior. When the similarity is greater than 0, it means that there is a business match between the current enterprise and the non-platform settled enterprise. S323. When the similarity is less than or equal to 0, the current electronic bill in the file is marked in blue, indicating that the current electronic bill is a questionable electronic bill, and further manual verification is performed afterwards.

[0024] It should be noted that by querying the related enterprises within the platform for non-platform enterprises, sorting them based on the degree of relevance and building business profiles of related enterprises, and combining the business profiles of related enterprises with the business profiles of current enterprises for judgment, the current electronic invoices are initially judged and marked for suspicion, thereby reducing the workload of manual verification; S4. When there are no related companies, collect the business scope of the company, combine the current business scope of the company to judge the suspicious bills, and mark the suspicious bills based on the judgment results; S41. When there are no affiliated enterprises outside the platform, the business scope of enterprises outside the platform is collected, and the cosine similarity is determined based on the actual business profile of the current enterprise. When the similarity value is less than or equal to 0, it means that there is no business intersection between the current enterprise and the enterprises outside the platform, and the current electronic invoice is a confirmed suspicious invoice, which is marked in red in the enterprise file. When the similarity value is greater than 0, it means that the business scope of the enterprises outside the platform overlaps with the actual business scope of the current enterprise, and the current electronic invoice is recorded as a suspicious electronic invoice, which is marked in blue in the enterprise file and awaits further manual verification.

[0025] It should be noted that the similarity comparison between non-platform companies and the actual business portraits of current companies based on business scope is relatively general. It is necessary to mark the bills with a similarity greater than 0 as suspicious and wait for further manual verification of the suspicion.

[0026] Example 3: Figure 2 As shown, an industry multi-dimensional data fusion system includes an enterprise data collection module, a business scope generation module, a suspicious degree determination module, and a manual verification module; The enterprise data collection module, based on the electronic invoice service platform, establishes enterprise files for enterprises within the platform through MySQL, collects electronic invoice information of enterprises within the platform, including bills of exchange, sales invoices, and purchase invoices, and records them in relevant enterprise files; The business scope generation module extracts bill data from different enterprise archives through OCR tools. The extracted bill commodity names are classified by similarity to generate actual business portraits of different enterprises; The suspiciousness judgment module retrieves the bill data under the archives of both companies for the companies registered in the platform, judges suspicious bills based on the overlap of business portraits, and marks the confirmed suspicious bills in red. For companies not registered in the platform, it retrieves all bill data stored in the platform, generates a list of related companies, generates a portrait of the related business scope based on the bill data in the archives of the related companies, judges suspicious bills and marks them in blue. When there are no related companies, it collects the business scope of the companies, combines the current business scope portrait of the companies, judges the suspicious bills, and marks the confirmed suspicious bills in red and the suspicious bills in blue based on the judgment results. The manual verification module conducts further manual verification of the suspicious bills marked in blue in the corporate archives to determine the suspicious degree of the electronic bills.

[0027] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for fusion of multi-dimensional industrial data, characterized in that: The method comprises the following steps: S1. Build an electronic invoice service platform, establish enterprise archives for enterprises within the platform based on MySQL, collect electronic invoice information of enterprises within the platform, including bills of exchange, sales invoices, and purchase invoices, and record them in relevant enterprise archives; S2. Extract information from bills based on the internal electronic bill data of the enterprise. For the enterprises settled in the platform, retrieve the bill data under the archives of both enterprises, establish business scope portraits of both enterprises, determine suspicious bills based on the overlap of business portraits, and mark suspicious bills; S3. For enterprises not registered in the platform, retrieve all bill data stored in the platform, generate a list of related enterprises, generate a business scope profile based on the bill data in the files of related enterprises, determine suspicious bills and mark suspicious bills based on the determination results; S4. When there are no affiliated companies, collect the business scope of the company, combine it with the current business scope portrait of the company, judge the suspicious bills, and mark the suspicious bills based on the judgment results.

2. The method for industrial multi-dimensional data fusion according to claim 1, characterized in that: The S2 comprises the following sub-steps: S21. Extracting bill information based on the internal electronic bill data of the enterprise archives. The specific steps are as follows: For bills of exchange in corporate archives, use OCR tools to search for keywords containing "payee" and "drawer" and extract the company name after the keywords; For sales invoices in corporate archives, use OCR tools to search for keywords containing "buyer" and "seller" and extract the company name after the keywords; For purchase invoices in corporate archives, use OCR tools to search for keywords containing "seller" and "buyer" and extract the company name after the keywords; S22. Based on the company names of both parties, retrieve the electronic invoice data under the company archives of both parties, build a portrait of the business scope of both companies, and determine the suspiciousness of the current electronic invoice based on the overlap of the business portraits.

3. The method for industrial multi-dimensional data fusion according to claim 2, characterized in that: The S22 comprises the following sub-steps: S221. Based on the names of both companies, search the company files of both companies in the platform, summarize the sales invoice data and purchase invoice data in the company files of both companies, extract the product name field based on the OCR tool, and The extracted product names are classified by similarity, and the algorithm formula is: ； in, is the similarity, Represents the similarity threshold, when ≥ When , the two product names are classified into one category. < When , it means that the two commodities are not of the same category; S222. Calculate the weight ratios of different commodity categories for the classified commodity categories in the sales invoices and purchase invoices, and establish the purchase business profile and sales business profile of Enterprise A and Enterprise B. The specific steps are as follows: For purchase invoices, for each commodity category , add up the amounts of all purchase invoices under the commodity category, and get the total amount of all purchase invoices under the commodity category , and add up the total amount of all purchase invoices to get the total purchase amount , calculate the purchase weight proportion of different commodity categories respectively, and the algorithm formula is: ； in, Representative product categories Procurement weight ratio under the current enterprise; For sales invoices, for each product category , add up the amounts of all sales invoices under the product category, and get the total amount of all sales invoices under the product category , and add up the total amount of all sales invoices to get the total sales amount , calculate the sales weight proportion of different commodity categories respectively, and the algorithm formula is: ； in, Representative product categories Based on the sales weight ratio of the current enterprise, establish product categories for Enterprise A and Enterprise B respectively. The steps of purchasing business portrait and sales business portrait are as follows: Company A: , ; Company B: , ; in, Represents the purchasing weight of enterprise A in commodity category 1, represents the sales weight of enterprise A in product category 1, represents the purchasing weight of enterprise B in commodity category 1, Represents the sales weight of enterprise B in commodity category 1; S223. Based on the type of the current electronic bill, the business overlap between the current enterprise A and enterprise B is calculated based on the cosine similarity method. The specific steps are as follows: When the electronic bill is a bill of exchange of enterprise A and the payee is enterprise B, or the electronic bill is a purchase invoice of enterprise A and the seller is enterprise B, the similarity between the purchase business profile of enterprise A and the sales business profile of enterprise B is calculated. The specific algorithm formula is: ； in, Represents the similarity of purchasing business portraits. When the purchase business of company A and the sales business of company B do not match in scope, there is a mismatched transaction behavior, that is, the current electronic bill is a definite suspicious bill. This means that the purchasing business of enterprise A and the sales business of enterprise B are matched and have a basis for business cooperation, that is, the current electronic bill is a normal bill; When the electronic bill is a bill of exchange of enterprise A and the drawer is enterprise B, or the electronic bill is a sales invoice of enterprise A and the purchaser is enterprise B, the similarity between the sales business profile of enterprise A and the procurement business profile of enterprise B is calculated. The specific algorithm formula is: ； in, Represents the similarity of sales business portraits. When the sales business of company A and the purchasing business of company B do not match in scope, there is a mismatched transaction behavior, that is, the current electronic bill is a definite suspicious bill. This means that the sales business of enterprise A and the purchasing business of enterprise B are matched and have a basis for business cooperation, that is, the current electronic bill is a normal bill; S224. For confirmed abnormal electronic invoices, the relevant enterprise files shall be marked in red, and the confirmed abnormal electronic invoices shall also be marked in red.

4. The method for industrial multi-dimensional data fusion according to claim 2, characterized in that: The S3 comprises the following steps: S31. Analyze the electronic invoice data according to the current enterprise electronic invoice type, the steps are as follows: When the enterprise information in the electronic invoice contains non-platform internal enterprises, extract the name of the non-platform internal enterprise , and retrieve all the bill data stored in the platform at the same time, and generate a list of related companies based on the names of non-platform companies; S32. Determine the suspiciousness of the current electronic invoice in combination with the list of related companies.

5. The method for industrial multi-dimensional data fusion according to claim 4, characterized in that: The S31 includes the following sub-steps: S311. Traverse the internal enterprise archives of the platform to obtain the data set of all enterprise bills stored in the platform , from the ticket data set Filter out the names of companies that are not on the platform Notes: ； in, express A specific electronic ticket in For non-platform internal enterprises A subset of electronic invoices, from Extraction and Companies that have business dealings with the platform , the steps are: ； in, Represents the collection of enterprise names that have settled in the platform. For A collection of companies that have business dealings with the platform; S312. For each enterprise settled in the platform , calculate its The specific steps are: ； in, Represents the co-occurrence frequency, that is, the Collection of bills middle Number of occurrences; ； in, Represents the total transaction amount, which is calculated in the and The sum of all transaction amounts in the bill is based on and Calculating association strength The algorithm formula is: ； in, Represents the weight coefficient, according to the strength of association Sort the internal affiliated enterprises of the platform that are not currently settled in the platform in descending order of strength, and select the top The companies that have settled in the platform serve as business profile comparison companies for current electronic invoice companies.

6. The method for industrial multi-dimensional data fusion according to claim 5, characterized in that: The S32 comprises the following steps: S321, based on the previous For each enterprise settled in the platform, the method of step S22 is adopted to obtain business portraits of the business portrait comparison enterprises respectively, and the business portraits of all the business portrait comparison enterprises are merged to obtain business portraits of the associated enterprises; S322. Calculate the similarity between the business profile of the current electronic invoice enterprise and the business profile of the associated enterprise based on the cosine similarity. When the similarity is less than or equal to 0, it means that there is a mismatch between the current enterprise and the non-platform settled enterprise in the transaction behavior. When the similarity is greater than 0, it means that there is a business match between the current enterprise and the non-platform settled enterprise. S323. When the similarity is less than or equal to 0, the current electronic bill in the file is marked in blue, indicating that the current electronic bill is a questionable electronic bill, and further manual verification is performed afterwards.

7. The method for industrial multi-dimensional data fusion according to claim 1, characterized in that: The S4 comprises the following sub-steps: S41. When there are no affiliated enterprises outside the platform, the business scope of enterprises outside the platform is collected, and the cosine similarity is determined based on the actual business profile of the current enterprise. When the similarity value is less than or equal to 0, it means that there is no business intersection between the current enterprise and the enterprises outside the platform, and the current electronic invoice is a confirmed suspicious invoice, which is marked in red in the enterprise file. When the similarity value is greater than 0, it means that the business scope of the enterprises outside the platform overlaps with the actual business scope of the current enterprise, and the current electronic invoice is recorded as a suspicious electronic invoice, which is marked in blue in the enterprise file and awaits further manual verification.

8. An industrial multi-dimensional data fusion system, characterized in that: The system adopts the industrial multi-dimensional data fusion method as described in any one of claims 1 to 7, including an enterprise data collection module, a business scope generation module, a suspicion determination module, and a manual verification module; The enterprise data collection module, based on the constructed electronic invoice service platform, establishes enterprise archives for enterprises within the platform through MySQL, collects electronic invoice information of enterprises within the platform, including bills of exchange, sales invoices, and purchase invoices, and records them in relevant enterprise archives; The business scope generation module extracts bill data based on bill data in different enterprise archives through OCR tools. The extracted bill commodity names are classified by similarity to generate actual business portraits of different enterprises; The suspiciousness determination module retrieves the bill data under the archives of both companies for the companies settled in the platform, determines suspicious bills based on the overlap of business portraits, and marks the confirmed suspicious bills in red. For companies not settled in the platform, it retrieves all bill data stored in the platform, generates a list of related companies, generates a portrait of the related business scope based on the bill data in the archives of the related companies, determines suspicious bills and marks them in blue. When there are no related companies, it collects the business scope of the companies, combines the current business scope portrait of the companies, determines the suspicious bills, and marks the confirmed suspicious bills in red and the suspicious bills in blue based on the determination results. The manual verification module further conducts manual verification on the suspicious bills marked in blue in the enterprise archives to determine the suspicious degree of the electronic bills.

Images