Data analysis system and method for financial automatic auditing platform

Abstract

The invention discloses a data analysis system and method for a financial automatic auditing platform, and relates to the technical field of financial platform data analysis. According to a work order declaration period, a period bill data set and a period work order text data set are generated; analyzing the periodic bill data set and the periodic work order text data set, identifying and marking abnormal data in the data sets, extracting the marked abnormal data, and forming an abnormal bill data set and an abnormal work order text data set; calculating a declaration abnormal value of each user in different work order declaration periods, and calculating an abnormal degree fluctuation value of each user according to the declaration abnormal value; analyzing an abnormal degree fluctuation mean value of the user according to the abnormal degree fluctuation value of the user, and performing security level evaluation on the user according to the abnormal degree fluctuation mean value; re-checking the bill images and the declaration work orders of the users of different grades by adopting different spot check proportions; the overall operation accuracy and efficiency of the system are improved; an accidental error caused by single calculation is prevented, and the robustness of the system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of financial platform data analysis, and in particular to a data analysis system and method for a financial automation audit platform. Background Art

[0002] With the continuous development and in-depth application of new-generation information technologies such as big data and artificial intelligence, the digitalization and intelligent construction of finance has become an extremely important part of the digital transformation process of today's enterprises.

[0003] As the scale of enterprises expands, financial accountants need to process and review a large number of financial documents every day. As the scale of the company's business continues to grow, more and more reimbursements, financial bills and business documents need to be processed; at the same time, the company's internal requirements for the refinement of financial management are getting higher and higher, and there is an urgent need to improve the efficiency and quality of financial work. The traditional model of relying on manual financial audits in the past has the disadvantages of low efficiency and high error rate.

[0004] Therefore, the present invention discloses a data analysis system and method for a financial automation audit platform to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide a data analysis system and method for a financial automation audit platform to solve the problems raised in the prior art.

[0006] To achieve the above object, the present invention provides the following technical solution: a data analysis method for a financial automation audit platform, the method comprising the following steps:

[0007] S1: Use image recognition technology to identify historical bill data from historical bill images; use text recognition technology to obtain the work order text data of historical declaration work orders; set the work order declaration cycle for the financial automation audit platform, and extract the historical bill data and work order text data of the financial automation audit platform according to the work order declaration cycle to generate a periodic bill data set and a periodic work order text data set;

[0008] S2: Analyze the periodic bill data set and the periodic work order text data set, identify and mark the abnormal data in the data set, extract the marked abnormal data, and form the abnormal bill data set and the abnormal work order text data set;

[0009] S3: Calculate the abnormal report value of each user in different work order reporting cycles based on the abnormal bill data set and the abnormal work order text data set, and calculate the abnormal degree fluctuation value of each user based on the abnormal report value;

[0010] S4: Analyze the user's abnormality fluctuation mean value based on the user's abnormality fluctuation value, and evaluate the user's security level based on the abnormality fluctuation mean value; use different sampling ratios to review the ticket images and declaration work orders of users of different levels.

[0011] According to the above scheme, S1 includes the following contents:

[0012] S101: Use image recognition technology to identify historical bill data from historical bill images; the historical bill data includes bill type, bill name, bill time, bill amount and bill personnel; use text recognition technology to obtain work order text data of historical declaration work orders, the work order text data includes declaration work order type, declaration work order name, declaration work order time, declaration work order amount and declaration work order personnel; uniformly record and number the users of the financial automation audit platform, and record the i-th user as U i , i∈[1,I], where I is the total number of users of the financial automation audit platform;

[0013] Based on image recognition technology and text recognition technology, it can quickly and accurately identify the data in historical bill images and historical declaration work orders, avoiding the inefficient working mode of relying on manual financial audits and the problem of manual errors; improving the overall operation accuracy and efficiency of the system;

[0014] S102: Set the work order declaration cycle for the financial automation audit platform, extract the data of the jth work order declaration cycle, generate a period bill data set and a period work order text data set; i The periodic bill data set generated by the historical bill data extracted in the jth work order declaration period is denoted as BD (i,j) ={BD (i,j) 1 , B.D. (i,j) 2 , ..., BD (i,j) Q}, and record the qth historical bill data as BD (i,j) q , q∈[1,Q], where Q represents the periodic bill data set BD (i,j) The number of historical ticket data in the i The periodic work order text data set generated by the historical work order extracted in the jth work order declaration cycle is denoted as TD (i,j) ={TD (i,j) 1 , TD (i,j) 2 , ..., TD (i,j) P}, and record the pth work order text data as TD (i,j) p , p∈[1,P]; where P represents the periodic work order text data set BD (i,j) The number of historical work orders reported; P>Q.

[0015] According to the above scheme, S2 includes the following contents:

[0016] S201: Associating the historical bill data and the work order text data belonging to the same historical declaration work order; marking the work order text data that does not have the associated corresponding historical bill data;

[0017] The work order declaration needs to be based on the actual bill. The work order text data that does not have the corresponding historical bill data is easy to be tampered by the user, which will cause losses. Therefore, marking the work order text data that does not have the corresponding historical bill data is conducive to avoiding losses and improving the overall accuracy of the system.

[0018] S202: Compare the historical ticket data and work order text data that have associated correspondence with the declaration rule base preset by the administrator, and mark the historical ticket data and work order text data that do not satisfy the declaration rule base; compare the historical ticket data and work order text data that have associated correspondence and satisfy the declaration rule base, and mark the historical ticket data and work order text data that are different;

[0019] S203: Extract the marked historical bill data and work order text data, and generate an abnormal bill data set and an abnormal work order text data set respectively.

[0020] Mark and extract a variety of data with abnormal conditions, providing data support for subsequent analysis and calculation;

[0021] According to the above scheme, S3 includes the following:

[0022] S301: Calculate the user U according to the abnormal bill data set and the abnormal work order text data set i The abnormal value of the report in the jth work order reporting cycle DO (i,j) , the specific calculation formula is as follows:

[0023] DO (i,j) =num 2 [ABD (i,j) ∪ATD (i,j) ]×(A×B) -1 ;

[0024] Among them, ABD (i,j) ∪ATD (i,j)Indicates that the abnormal bill data set ABD (i,j) and abnormal work order text data set ATD (i,j) Perform union processing to merge the historical ticket data and work order text data that have an associated corresponding relationship into one element;

[0025] num 2 [ABD (i,j) ∪ATD (i,j) ] represents the abnormal bill data set ABD (i,j) and abnormal work order text data set ATD (i,j) The square of the number of historical work orders involved in the union set. A represents the abnormal bill data set ABD (i,j) The number of historical ticket images involved in the historical ticket data, B represents the abnormal work order text data set ATD (i,j) The number of historical work orders involved in the work order text data;

[0026] S302: According to user U i The abnormal value of the report in each work order reporting cycle is calculated by the user U i Abnormal degree fluctuation value IDO i ; The specific calculation formula is as follows:

[0027] IDO i =C1(A÷Q+C2×B÷P)×∑ j J ＝1 {[DO (i,j) -u] 2 ÷σ 2};

[0028] Among them, C1 and C2 represent the fluctuation coefficients, which are preset constants; u represents the abnormal value DO reported (i,j) The mean of ,σ represents the reported outlier value DO (i,j) The standard deviation of; J represents the total number of work order reporting cycles.

[0029] Calculating the abnormal degree fluctuation value for multiple work order reporting cycles can avoid occasional errors caused by abnormal reporting values ​​in a single work order reporting cycle and improve the overall accuracy of the system;

[0030] According to the above scheme, S4 includes the following contents:

[0031] S401: Based on J work order reporting cycles, randomly remove one work order reporting cycle; repeat S301 and S302 based on the remaining J-1 work order reporting cycles, and calculate the abnormal degree fluctuation value recorded as IDO i J-1; Based on the remaining J-1 work order reporting cycles, randomly remove another work order reporting cycle; repeat S301 and S302 based on the remaining J-2 work order reporting cycles, and calculate the abnormal degree fluctuation value and record it as IDO i J-2 ; Calculate user U i Abnormal degree fluctuation value IDO i 、IDO i J-1 and IDO i J-2 The mean of is recorded as the mean of abnormal degree fluctuation;

[0032] By randomly removing a work order declaration cycle multiple times, calculating the abnormality fluctuation value again, and calculating the abnormality fluctuation mean based on multiple abnormality fluctuation values, the accuracy of subsequent safety level division can be further guaranteed, accidental errors caused by a single calculation can be prevented, and the robustness of the system can be improved;

[0033] S402: Setting a first threshold and a second threshold for the abnormal degree fluctuation mean, wherein the first threshold is greater than the second threshold; if user U i If the average fluctuation of abnormal degree is greater than or equal to the first threshold, user U i Recorded as the first security level user; if user U i If the average abnormality fluctuation of user U is less than the first threshold and greater than or equal to the second threshold, user U i Recorded as the second security level user; if user U i If the average fluctuation of abnormal degree is less than the second threshold, user U i Recorded as the third security level user;

[0034] S403: The real-time bill images and application work orders of the third security level users are subject to system automated review. By random sampling, some bill images and application work orders that have passed the automated review are subject to simple manual review on a regular basis. Simple manual review means that only the risk field, amount and date are checked. The risk field is preset by the system.

[0035] During the bill review process, risk fields are usually key fields that are prone to false, fraudulent or non-compliant situations; risk fields include office expenses, other expenses, miscellaneous expenses and non-compliant supplier names;

[0036] Reimbursement of items with uncommon or vague names is prone to false information; vague names include office expenses, other expenses and miscellaneous expenses; non-compliant supplier names include the names of merchants that are not on the company's supplier list or blacklist.

[0037] The real-time ticket images and application work orders of users at the second security level shall first be subject to system automated review. After the automated review is passed, all real-time ticket images and application work orders shall be subject to simple manual review. The ticket images and application work orders of users at the second security level shall be subject to detailed review on a regular basis through random sampling. Detailed review means checking all data of the ticket images and application work orders. The sampling ratio of users at the second security level is higher than that of users at the third security level. The sampling ratio is equal to the sum of the number of ticket images and application work orders sampled divided by the sum of the total number of historical ticket images and historical application work orders.

[0038] The real-time ticket images and declaration work orders of users with the first security level will first be subject to system automated review. After the automated review is passed, all real-time ticket images and declaration work orders will need to be manually reviewed in detail. The ticket images and declaration work orders of users with the first security level will be reviewed in detail regularly through random inspections. The inspection ratio for users with the first security level is higher than that for users with the second security level.

[0039] Another aspect of the present application provides a data analysis system for a financial automation audit platform, the system is applied to the implementation of the data analysis method for a financial automation audit platform, the system includes a data identification and integration module, a data analysis and splitting module, an abnormal data calculation module and a user level assessment and audit module;

[0040] The data recognition and integration module is used to use image recognition technology to perform historical bill data recognition on historical bill images; use text recognition technology to obtain the work order text data of historical declaration work orders; set the work order declaration cycle for the financial automation audit platform, and extract the historical bill data and work order text data of the financial automation audit platform according to the work order declaration cycle to generate a periodic bill data set and a periodic work order text data set;

[0041] The data analysis and splitting module is used to analyze the periodic bill data set and the periodic work order text data set, identify and mark abnormal data in the data set, extract the marked abnormal data, and form an abnormal bill data set and an abnormal work order text data set;

[0042] The abnormal data calculation module calculates the abnormal declaration value of each user in different work order declaration cycles according to the abnormal bill data set and the abnormal work order text data set, and calculates the abnormal degree fluctuation value of each user according to the abnormal declaration value;

[0043] The user level assessment and audit module is used to analyze the user's abnormality degree fluctuation mean value according to the user's abnormality degree fluctuation value, and evaluate the user's security level according to the abnormality degree fluctuation mean value; and use different sampling ratios to review the ticket images and declaration work orders of users of different levels.

[0044] According to the above solution, the data identification and integration module includes a data identification unit and a periodic data integration unit;

[0045] The data recognition unit uses image recognition technology to perform historical bill data recognition on historical bill images; the historical bill data includes bill type, bill name, bill time, bill amount and bill personnel; the text recognition technology is used to obtain the work order text data of the historical declaration work order, and the work order text data includes the declaration work order type, declaration work order name, declaration work order time, declaration work order amount and declaration work order personnel;

[0046] The cycle data integration unit is used to set a work order declaration cycle for the financial automation audit platform, extract data according to the work order declaration cycle, and generate a cycle bill data set and a cycle work order text data set.

[0047] According to the above scheme, the data analysis and splitting module includes a data association unit and an abnormal mark extraction unit;

[0048] The data association unit is used to associate the historical ticket data and the work order text data belonging to the same historical declaration work order;

[0049] The abnormal mark extraction unit is used to mark the work order text data of the historical ticket data that does not have an associated corresponding data; compare the historical ticket data and work order text data that have an associated corresponding data with the declaration rule base preset by the administrator, and mark the historical ticket data and work order text data that do not satisfy the declaration rule base; compare the historical ticket data and work order text data that have an associated corresponding data and satisfy the declaration rule base, and mark the historical ticket data and work order text data that have a different difference; extract the marked historical ticket data and work order text data, and generate an abnormal ticket data set and an abnormal work order text data set respectively.

[0050] According to the above scheme, the abnormal data calculation module is used to declare the abnormal calculation unit and the abnormal degree fluctuation unit;

[0051] The abnormal declaration calculation unit is used to calculate the abnormal declaration value of each user in each work order declaration cycle according to the abnormal bill data set and the abnormal work order text data set;

[0052] The abnormality degree fluctuation unit is used to calculate the abnormality degree fluctuation value of the user according to the abnormality value reported by the user in each work order reporting cycle.

[0053] According to the above scheme, the user level assessment and review module includes an abnormal degree fluctuation mean value calculation unit and a level division and review unit;

[0054] The abnormality degree fluctuation mean value calculation unit is used to randomly remove a work order declaration cycle based on the number of original work order declaration cycles; calculate the abnormality degree fluctuation value based on the remaining work order declaration cycles; randomly remove a work order declaration cycle again based on the remaining work order declaration cycles; calculate the abnormality degree fluctuation mean value based on the remaining work order declaration cycles; calculate the abnormality degree fluctuation mean value of the user;

[0055] The grading review unit is used to set a first threshold and a second threshold for the abnormality degree fluctuation mean, the first threshold is greater than the second threshold; if the abnormality degree fluctuation mean of the user is greater than or equal to the first threshold, the user is recorded as a first security level user; if the abnormality degree fluctuation mean of the user is less than the first threshold and greater than or equal to the second threshold, the user is recorded as a second security level user; if the abnormality degree fluctuation mean of the user is less than the second threshold, the user is recorded as a third security level user; different sampling ratios are used to review the ticket images and declaration work orders of users of different levels.

[0056] Compared with the prior art, the beneficial effects of the present invention are as follows: based on image recognition technology and text recognition technology, it is capable of quickly and accurately identifying data in historical bill images and historical declaration work orders, thus avoiding the inefficient working mode of relying on manual financial audits and the problem of manual error; improving the overall operation accuracy and efficiency of the system; the declaration work order needs to be based on the actual bill, and the work order text data that does not have the corresponding historical bill data is easily tampered by the user, thereby causing losses; therefore, marking the work order text data that does not have the corresponding historical bill data is beneficial to avoiding losses and improving the overall accuracy of the system; marking and extracting a variety of data with abnormal conditions, providing data support for subsequent analysis and calculation; by randomly removing a work order declaration cycle multiple times, calculating the abnormality degree fluctuation value again, and calculating the abnormality degree fluctuation mean based on multiple abnormality degree fluctuation values, which can further ensure the accuracy of subsequent security level division, prevent occasional errors caused by a single calculation, and improve the robustness of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0058] Figure 1 It is a flow chart of the data analysis method for the financial automation audit platform of the present invention;

[0059] Figure 2 The present invention is a schematic diagram of the structure of the data analysis system used for the financial automation audit platform. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] See also Figure 1 The present invention provides a technical solution: a data analysis method for a financial automation audit platform, the method comprising the following steps:

[0062] S1: Use image recognition technology to identify historical bill data from historical bill images; use text recognition technology to obtain the work order text data of historical declaration work orders; set the work order declaration cycle for the financial automation audit platform, and extract the historical bill data and work order text data of the financial automation audit platform according to the work order declaration cycle to generate a periodic bill data set and a periodic work order text data set;

[0063] In S1, the following are included:

[0064] S101: Use image recognition technology to identify historical bill data from historical bill images; historical bill data includes bill type, bill name, bill time, bill amount and bill personnel; use text recognition technology to obtain work order text data of historical declaration work orders, and the work order text data includes declaration work order type, declaration work order name, declaration work order time, declaration work order amount and declaration work order personnel; uniformly record and number the users of the financial automation audit platform, and record the i-th user as U i , i∈[1,I], where I is the total number of users of the financial automation audit platform;

[0065] S102: Set the work order declaration cycle for the financial automation audit platform, extract the data of the jth work order declaration cycle, generate a period bill data set and a period work order text data set; i The periodic bill data set generated by the historical bill data extracted in the jth work order declaration period is denoted as BD (i,j) ={BD (i,j) 1 , B.D. (i,j) 2 , ..., BD (i,j) Q}, and record the qth historical bill data as BD (i,j) q , q∈[1,Q], where Q represents the periodic bill data set BD (i,j)The number of historical ticket data in the i The periodic work order text data set generated by the historical work order extracted in the jth work order declaration cycle is denoted as TD (i,j) ={TD (i,j) 1 , TD (i,j) 2 , ..., TD (i,j) P}, and record the pth work order text data as TD (i,j) p , p∈[1,P]; where P represents the periodic work order text data set BD (i,j) The number of historical work orders reported; P>Q.

[0066] S2: Analyze the periodic bill data set and the periodic work order text data set, identify and mark the abnormal data in the data set, extract the marked abnormal data, and form the abnormal bill data set and the abnormal work order text data set;

[0067] In S2, the following are included:

[0068] S201: Associating the historical bill data and the work order text data belonging to the same historical declaration work order; marking the work order text data that does not have the associated corresponding historical bill data;

[0069] S202: Compare the historical ticket data and work order text data that have associated correspondence with the declaration rule base preset by the administrator, and mark the historical ticket data and work order text data that do not satisfy the declaration rule base; compare the historical ticket data and work order text data that have associated correspondence and satisfy the declaration rule base, and mark the historical ticket data and work order text data that are different;

[0070] S203: Extract the marked historical bill data and work order text data, and generate an abnormal bill data set and an abnormal work order text data set respectively.

[0071] Example 1: In this example, the declaration rule base preset by the administrator has provisions for the reimbursement amount of business trip travel expenses: for the business trip point Z city, the maximum accommodation amount per person per night is Y yuan;

[0072] If user U i In the historical work order, the work order name is the two-night business trip accommodation fee for the individual; the work order amount is (2Y+10) yuan;

[0073] Fixed User iThe accommodation fee for a single person on a business trip is (Y+5) yuan per night, which exceeds the provisions of the declaration rule library preset by the administrator regarding the reimbursement amount of business trip travel expenses; in this embodiment 1, user U i Mark the work order text data;

[0074] Example 2: In this example, the declaration rule base preset by the administrator has provisions for the reimbursement amount of business trip travel expenses: for the business trip point Z city, the maximum accommodation amount per person per night is Y yuan;

[0075] If user U i In the historical work order, the work order name is the two-night business trip accommodation fee for the individual; the work order amount is (2Y-10) yuan;

[0076] In the historical bills associated with the historical application work orders, the bill name is the two-night business trip accommodation fee for the applicant; the bill amount is (2Y-20) yuan;

[0077] The amount filled in by the user in the historical application work order and the historical bill is inconsistent. In this embodiment 2, user U i Mark the work order text data and historical ticket data;

[0078] S3: Calculate the abnormal report value of each user in different work order reporting cycles based on the abnormal bill data set and the abnormal work order text data set, and calculate the abnormal degree fluctuation value of each user based on the abnormal report value;

[0079] In S3, include the following:

[0080] S301: Calculate the user U according to the abnormal bill data set and the abnormal work order text data set i The abnormal value of the report in the jth work order reporting cycle DO (i,j) , the specific calculation formula is as follows:

[0081] DO (i,j) =num 2 [ABD (i,j) ∪ATD (i,j) ]×(A×B) -1 ;

[0082] Among them, ABD (i,j) ∪ATD (i,j) Indicates that the abnormal bill data set ABD (i,j) and abnormal work order text data set ATD (i,j) Perform union processing to merge the historical ticket data and work order text data that have an associated corresponding relationship into one element;

[0083] num 2 [ABD(i,j) ∪ATD (i,j) ] represents the abnormal bill data set ABD (i,j) and abnormal work order text data set ATD (i,j) The square of the number of historical work orders involved in the union set. A represents the abnormal bill data set ABD (i,j) The number of historical ticket images involved in the historical ticket data, B represents the abnormal work order text data set ATD (i,j) The number of historical work orders involved in the work order text data;

[0084] S302: According to user U i The abnormal value of the report in each work order reporting cycle is calculated by the user U i Abnormal degree fluctuation value IDO i ; The specific calculation formula is as follows:

[0085] IDO i =C1(A÷Q+C2×B÷P)×∑ j J ＝1 {[DO (i,j) -u] 2 ÷σ 2};

[0086] Among them, C1 and C2 represent the fluctuation coefficients, which are preset constants; u represents the abnormal value DO reported (i,j) The mean of ,σ represents the reported outlier value DO (i,j) The standard deviation of; J represents the total number of work order reporting cycles.

[0087] S4: Analyze the user's abnormality fluctuation mean value based on the user's abnormality fluctuation value, and evaluate the user's security level based on the abnormality fluctuation mean value; use different sampling ratios to review the ticket images and declaration work orders of users of different levels.

[0088] In S4, the following are included:

[0089] S401: Based on J work order reporting cycles, randomly remove one work order reporting cycle; repeat S301 and S302 based on the remaining J-1 work order reporting cycles, and calculate the abnormal degree fluctuation value recorded as IDO i J-1 ; Based on the remaining J-1 work order reporting cycles, randomly remove another work order reporting cycle; repeat S301 and S302 based on the remaining J-2 work order reporting cycles, and calculate the abnormal degree fluctuation value and record it as IDO i J-2 ; Calculate user U i Abnormal degree fluctuation value IDOi 、IDO i J-1 and IDO i J-2 The mean of is recorded as the mean of abnormal degree fluctuation;

[0090] S402: Setting a first threshold and a second threshold for the abnormal degree fluctuation mean, wherein the first threshold is greater than the second threshold; if user U i If the average fluctuation of abnormal degree is greater than or equal to the first threshold, user U i Recorded as the first security level user; if user U i If the average abnormality fluctuation of user U is less than the first threshold and greater than or equal to the second threshold, user U i Recorded as the second security level user; if user U i If the average fluctuation of abnormal degree is less than the second threshold, user U i Recorded as the third security level user;

[0091] S403: The real-time bill images and application work orders of the third security level users are subject to system automated review. By random sampling, some bill images and application work orders that have passed the automated review are subject to simple manual review on a regular basis. Simple manual review means that only the risk field, amount and date are checked. The risk field is preset by the system.

[0092] The real-time ticket images and application work orders of users at the second security level shall first be subject to system automated review. After the automated review is passed, all real-time ticket images and application work orders shall be subject to simple manual review. The ticket images and application work orders of users at the second security level shall be subject to detailed review on a regular basis through random sampling. Detailed review means checking all data of the ticket images and application work orders. The sampling ratio of users at the second security level is higher than that of users at the third security level. The sampling ratio is equal to the sum of the number of ticket images and application work orders sampled divided by the sum of the total number of historical ticket images and historical application work orders.

[0093] The real-time ticket images and declaration work orders of users with the first security level will first be subject to system automated review. After the automated review is passed, all real-time ticket images and declaration work orders will need to be manually reviewed in detail. The ticket images and declaration work orders of users with the first security level will be reviewed in detail regularly through random inspections. The inspection ratio for users with the first security level is higher than that for users with the second security level.

[0094] See also Figure 2 , the present invention provides a technical solution: a data analysis system for a financial automation audit platform, the system comprising a data identification and integration module, a data analysis and splitting module, an abnormal data calculation module and a user level assessment and audit module;

[0095] The data recognition and integration module is used to use image recognition technology to perform historical bill data recognition on historical bill images; use text recognition technology to obtain the work order text data of historical declaration work orders; set the work order declaration cycle for the financial automation audit platform, and extract the historical bill data and work order text data of the financial automation audit platform according to the work order declaration cycle to generate a periodic bill data set and a periodic work order text data set;

[0096] The data analysis and splitting module is used to analyze the periodic bill data set and the periodic work order text data set, identify and mark abnormal data in the data set, extract the marked abnormal data, and form an abnormal bill data set and an abnormal work order text data set;

[0097] The abnormal data calculation module calculates the abnormal value of each user's declaration in different work order declaration cycles based on the abnormal bill data set and the abnormal work order text data set, and calculates the abnormal degree fluctuation value of each user based on the declared abnormal value;

[0098] The user level assessment and review module is used to analyze the user's abnormality fluctuation mean based on the user's abnormality fluctuation value, and to assess the user's security level based on the abnormality fluctuation mean; different sampling ratios are used to review the ticket images and declaration work orders of users of different levels.

[0099] The data identification and integration module includes a data identification unit and a periodic data integration unit;

[0100] The data recognition unit uses image recognition technology to identify historical bill data from historical bill images; the historical bill data includes bill type, bill name, bill time, bill amount and bill personnel; the text recognition technology is used to obtain the work order text data of historical declaration work orders, and the work order text data includes the declaration work order type, declaration work order name, declaration work order time, declaration work order amount and declaration work order personnel;

[0101] The cycle data integration unit is used to set the work order reporting cycle for the financial automation audit platform, extract data according to the work order reporting cycle, and generate a cycle bill data set and a cycle work order text data set.

[0102] The data analysis and splitting module includes a data association unit and an abnormal mark extraction unit;

[0103] The data association unit is used to associate the historical ticket data and the work order text data belonging to the same historical declaration work order;

[0104] The abnormal marking extraction unit is used to mark the work order text data of the historical ticket data that does not have an associated corresponding data; compare the historical ticket data and work order text data that have an associated corresponding data with the declaration rule base preset by the administrator, and mark the historical ticket data and work order text data that do not satisfy the declaration rule base; compare the historical ticket data and work order text data that have an associated corresponding data and satisfy the declaration rule base, and mark the historical ticket data and work order text data that have differences; extract the marked historical ticket data and work order text data, and generate an abnormal ticket data set and an abnormal work order text data set respectively.

[0105] The abnormal data calculation module is used to declare abnormal calculation units and abnormal degree fluctuation units;

[0106] The abnormal declaration calculation unit is used to calculate the abnormal declaration value of each user in each work order declaration cycle according to the abnormal bill data set and the abnormal work order text data set;

[0107] The abnormality degree fluctuation unit is used to calculate the abnormality degree fluctuation value of the user according to the abnormality value reported by the user in each work order reporting cycle.

[0108] The user level assessment and review module includes an abnormal degree fluctuation mean value calculation unit and a level classification and review unit;

[0109] The abnormality degree fluctuation mean value calculation unit is used to randomly remove a work order reporting cycle based on the number of original work order reporting cycles; calculate the abnormality degree fluctuation value based on the remaining work order reporting cycles; randomly remove a work order reporting cycle again based on the remaining work order reporting cycles; calculate the abnormality degree fluctuation mean value of the user;

[0110] The grading review unit is used to set a first threshold and a second threshold for the mean value of abnormality fluctuation, the first threshold being greater than the second threshold; if the mean value of the abnormality fluctuation of the user is greater than or equal to the first threshold, the user is recorded as a user of the first security level; if the mean value of the abnormality fluctuation of the user is less than the first threshold and greater than or equal to the second threshold, the user is recorded as a user of the second security level; if the mean value of the abnormality fluctuation of the user is less than the second threshold, the user is recorded as a user of the third security level; different sampling ratios are used to review the ticket images and declaration work orders of users of different levels.

[0111] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0112] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A data analysis method for a financial automation audit platform, characterized in that: The method comprises the following steps: S1: Use image recognition technology to identify historical bill data from historical bill images; use text recognition technology to obtain the work order text data of historical declaration work orders; set the work order declaration cycle for the financial automation audit platform, and extract the historical bill data and work order text data of the financial automation audit platform according to the work order declaration cycle to generate a periodic bill data set and a periodic work order text data set; S2: Analyze the periodic bill data set and the periodic work order text data set, identify and mark the abnormal data in the data set, extract the marked abnormal data, and form the abnormal bill data set and the abnormal work order text data set; S3: Calculate the abnormal report value of each user in different work order reporting cycles based on the abnormal bill data set and the abnormal work order text data set, and calculate the abnormal degree fluctuation value of each user based on the abnormal report value; S4: Analyze the user's abnormality fluctuation mean according to the user's abnormality fluctuation value, and evaluate the user's security level according to the abnormality fluctuation mean; Different sampling ratios are used to review the ticket images and declaration work orders of users of different levels.

2. The data analysis method for a financial automation audit platform according to claim 1, characterized in that: In S1, the following are included: S101: Use image recognition technology to identify historical bill data from historical bill images; the historical bill data includes bill type, bill name, bill time, bill amount and bill personnel; use text recognition technology to obtain work order text data of historical declaration work orders, the work order text data includes declaration work order type, declaration work order name, declaration work order time, declaration work order amount and declaration work order personnel; uniformly record and number the users of the financial automation audit platform, and record the i-th user as U i , i∈[1,I], where I is the total number of users of the financial automation audit platform; S102: Set the work order declaration cycle for the financial automation audit platform, extract the data of the jth work order declaration cycle, generate a period bill data set and a period work order text data set; i The periodic bill data set generated by the historical bill data extracted in the jth work order declaration period is denoted as BD (i,j) ={BD (i,j) 1 , B.D. (i,j) 2 , ..., BD (i,j) Q }, and record the qth historical bill data as BD (i,j) q , q∈[1,Q], where Q represents the periodic bill data set BD (i,j) The number of historical ticket data in the i The periodic work order text data set generated by the historical work order extracted in the jth work order declaration cycle is denoted as TD (i,j) ={TD (i,j) 1 , TD (i,j) 2 , ..., TD (i,j) P }, and record the pth work order text data as TD (i,j) p , p∈[1,P]; where P represents the periodic work order text data set BD (i,j) The number of historical work orders reported; P>Q.

3. The data analysis method for a financial automation audit platform according to claim 2, characterized in that: In S2, the following are included: S201: Associating the historical bill data and the work order text data belonging to the same historical declaration work order; marking the work order text data that does not have the associated corresponding historical bill data; S202: Compare the historical ticket data and work order text data that have associated correspondence with the declaration rule base preset by the administrator, and mark the historical ticket data and work order text data that do not satisfy the declaration rule base; compare the historical ticket data and work order text data that have associated correspondence and satisfy the declaration rule base, and mark the historical ticket data and work order text data that are different; S203: Extract the marked historical bill data and work order text data, and generate an abnormal bill data set and an abnormal work order text data set respectively.

4. The data analysis method for a financial automation audit platform according to claim 3 is characterized in that: In S3, include the following: S301: Calculate the user U according to the abnormal bill data set and the abnormal work order text data set i The abnormal value of the report in the jth work order reporting cycle DO (i,j) , the specific calculation formula is as follows: DO (i,j) =number 2 [USA (i,j) ∪ATD (i,j) ]×(A×B) -1 ; Among them, ABD (i,j) ∪ATD (i,j) Indicates that the abnormal bill data set ABD (i,j) and abnormal work order text data set ATD (i,j) Perform union processing to merge the historical ticket data and work order text data that have an associated corresponding relationship into one element; num 2 [ABD (i,j) ∪ATD (i,j) ] represents the abnormal bill data set ABD (i,j) and abnormal work order text data set ATD (i,j) The square of the number of historical work orders involved in the union set. A represents the abnormal bill data set ABD (i,j) The number of historical ticket images involved in the historical ticket data, B represents the abnormal work order text data set ATD (i,j) The number of historical work orders involved in the work order text data; S302: According to user U i The abnormal value of the report in each work order reporting cycle is calculated by the user U i Abnormal degree fluctuation value IDO i ; The specific calculation formula is as follows: Among them, C1 and C2 represent the fluctuation coefficients, which are preset constants; u represents the abnormal value DO reported (i,j) The mean of ,σ represents the reported outlier DO (i,j) The standard deviation of; J represents the total number of work order reporting cycles.

5. The data analysis method for a financial automation audit platform according to claim 4, characterized in that: In S4, the following are included: S401: Based on J work order reporting cycles, randomly remove one work order reporting cycle; repeat S301 and S302 based on the remaining J-1 work order reporting cycles, and calculate the abnormal degree fluctuation value recorded as IDO i J-1 ; Based on the remaining J-1 work order reporting cycles, randomly remove another work order reporting cycle; repeat S301 and S302 based on the remaining J-2 work order reporting cycles, and calculate the abnormal degree fluctuation value and record it as IDO i J-2 ; Calculate user U i Abnormal degree fluctuation value IDO i 、IDO i J-1 and IDO i J-2 The mean of is recorded as the mean of abnormal degree fluctuation; S402: Setting a first threshold and a second threshold for the abnormal degree fluctuation mean, wherein the first threshold is greater than the second threshold; if user U i If the average abnormal fluctuation of the user U is greater than or equal to the first threshold, the user U i Recorded as the first security level user; if user U i If the average abnormality fluctuation of user U is less than the first threshold and greater than or equal to the second threshold, user U i Recorded as the second security level user; if user U i If the average fluctuation of abnormal degree is less than the second threshold, user U i Recorded as the third security level user; S403: The real-time bill images and application work orders of the third security level users are subject to system automated review. By random sampling, some bill images and application work orders that have passed the automated review are subject to simple manual review on a regular basis. Simple manual review means that only the risk field, amount and date are checked. The risk field is preset by the system. The real-time ticket images and application work orders of users at the second security level shall first be subject to system automated review. After the automated review is passed, all real-time ticket images and application work orders shall be subject to simple manual review. The ticket images and application work orders of users at the second security level shall be subject to detailed review on a regular basis through random sampling. Detailed review means checking all data of the ticket images and application work orders. The sampling ratio of users at the second security level is higher than that of users at the third security level. The sampling ratio is equal to the sum of the number of ticket images and application work orders sampled divided by the sum of the total number of historical ticket images and historical application work orders. The real-time ticket images and declaration work orders of users with the first security level will first be subject to system automated review. After the automated review is passed, all real-time ticket images and declaration work orders will need to be manually reviewed in detail. The ticket images and declaration work orders of users with the first security level will be reviewed in detail regularly through random inspections. The inspection ratio for users with the first security level is higher than that for users with the second security level.

6. A data analysis system for a financial automation audit platform, the system being applied to the data analysis method for a financial automation audit platform as claimed in any one of claims 1 to 5, characterized in that: The system includes a data identification and integration module, a data analysis and splitting module, an abnormal data calculation module, and a user level assessment and review module; The data recognition and integration module is used to use image recognition technology to perform historical bill data recognition on historical bill images; use text recognition technology to obtain the work order text data of historical declaration work orders; set the work order declaration cycle for the financial automation audit platform, and extract the historical bill data and work order text data of the financial automation audit platform according to the work order declaration cycle to generate a periodic bill data set and a periodic work order text data set; The data analysis and splitting module is used to analyze the periodic bill data set and the periodic work order text data set, identify and mark abnormal data in the data set, extract the marked abnormal data, and form an abnormal bill data set and an abnormal work order text data set; The abnormal data calculation module calculates the abnormal declaration value of each user in different work order declaration cycles according to the abnormal bill data set and the abnormal work order text data set, and calculates the abnormal degree fluctuation value of each user according to the abnormal declaration value; The user level assessment and audit module is used to analyze the user's abnormal degree fluctuation mean value according to the user's abnormal degree fluctuation value, and conduct a security level assessment on the user according to the abnormal degree fluctuation mean value; Different sampling ratios are used to review the ticket images and declaration work orders of users of different levels.

7. The data analysis system for the financial automation audit platform according to claim 6, characterized in that: The data identification and integration module includes a data identification unit and a periodic data integration unit; The data recognition unit uses image recognition technology to perform historical bill data recognition on historical bill images; the historical bill data includes bill type, bill name, bill time, bill amount and bill personnel; the text recognition technology is used to obtain the work order text data of the historical declaration work order, and the work order text data includes the declaration work order type, declaration work order name, declaration work order time, declaration work order amount and declaration work order personnel; The cycle data integration unit is used to set a work order declaration cycle for the financial automation audit platform, extract data according to the work order declaration cycle, and generate a cycle bill data set and a cycle work order text data set.

8. The data analysis system for the financial automation audit platform according to claim 6, characterized in that: The data analysis and splitting module includes a data association unit and an abnormal mark extraction unit; The data association unit is used to associate the historical ticket data and the work order text data belonging to the same historical declaration work order; The abnormal marking extraction unit is used to mark the work order text data for which there is no associated historical ticket data; compare the associated historical ticket data and work order text data with the declaration rule base preset by the administrator, and mark the historical ticket data and work order text data that do not satisfy the declaration rule base; compare the historical ticket data and work order text data that have an associated correspondence and satisfy the declaration rule base, and mark the historical ticket data and work order text data that are different; extract the marked historical ticket data and work order text data, and generate an abnormal ticket data set and an abnormal work order text data set, respectively.

9. The data analysis system for the financial automation audit platform according to claim 6, characterized in that: The abnormal data calculation module is used to declare an abnormal calculation unit and an abnormal degree fluctuation unit; The abnormal declaration calculation unit is used to calculate the abnormal declaration value of each user in each work order declaration cycle according to the abnormal bill data set and the abnormal work order text data set; The abnormality degree fluctuation unit is used to calculate the abnormality degree fluctuation value of the user according to the abnormality value reported by the user in each work order reporting cycle.

10. The data analysis system for the financial automation audit platform according to claim 6, characterized in that: The user grade assessment and review module includes an abnormal degree fluctuation mean value calculation unit and a grade division review unit; The abnormality degree fluctuation mean value calculation unit is used to randomly remove a work order declaration cycle based on the number of original work order declaration cycles; and calculate the abnormality degree fluctuation value based on the remaining work order declaration cycles; Based on the remaining work order reporting cycles, a work order reporting cycle is randomly removed again; based on the remaining work order reporting cycles, the abnormality degree fluctuation value is calculated; Calculate the average fluctuation of the user's abnormality level; The level division review unit is used to set a first threshold and a second threshold for the abnormal degree fluctuation mean, and the first threshold is greater than the second threshold; if the abnormal degree fluctuation mean of the user is greater than or equal to the first threshold, the user is recorded as a first security level user; If the user's abnormal degree fluctuation mean is less than the first threshold and greater than or equal to the second threshold, the user is recorded as a second security level user; If the user's abnormal degree fluctuation mean is less than the second threshold, the user is recorded as a third security level user; Different sampling ratios are used to review the ticket images and declaration work orders of users of different levels.

Images