Intelligent student subsidy application auditing and approving system based on big data

Abstract

The invention discloses an intelligent student subsidy application auditing and approving system based on big data, and relates to the technical field of information processing management, and the system comprises the steps: obtaining a student address information recognition result through an information recognition module; the address information processing module splits and analyzes the address information to form standard address information, judges the hierarchical state of the standard address information, establishes an address hierarchical fixed matching detection model, and generates a processing signal; the address bidirectional matching module counts and analyzes all address abnormal signals, identifies the uppermost and lowermost addresses of abnormal address information, and generates a search result in combination with a preset forward / reverse derivation rule; the abnormal address information processing and decision-making module compares the consistency of each search result of each abnormal address information, applies a corresponding processing strategy, and further decides whether to execute automatic correction or abnormal prompt modification through a special situation judgment unit, so that automatic auditing and accurate processing of application information are realized, and the application information is more accurate. And the subsidy approval efficiency and accuracy are obviously improved.

Description

Technical Field

[0001] The present invention relates to the field of information processing management technology, and in particular to an intelligent student aid application review and approval system based on big data. Background Art

[0002] In the current student aid application review and approval system, image scanning technology is typically used to capture the information on the aid application form, and then optical character recognition (OCR) technology is used to extract the text information for further processing and analysis. However, due to the inaccuracy of image scanning, it is easy for student address information to be incomplete or incorrectly recognized. Once this happens, the system will usually return the application form to the applicant for self-revision, seriously affecting the efficiency of the review and approval process. This process not only wastes a lot of human resources, but also is not conducive to improving the timeliness and accuracy of approval.

[0003] Existing technologies usually directly return financial aid application forms containing incorrectly identified information to students, who are then required to correct the information themselves and submit it again. This approach not only increases the complexity for students to modify information, but also easily causes students to distrust the review and approval system, affecting the overall effectiveness of the financial aid application. Therefore, there is a need for an intelligent student financial aid application review and approval system that can automatically identify and correct address information errors and efficiently handle abnormal information. Summary of the Invention

[0004] The purpose of the present invention is to solve the problems in the background technology and to propose an intelligent student aid application review and approval system based on big data.

[0005] In order to achieve the above objectives, the present invention adopts the following technical solutions: an intelligent student aid application review and approval system based on big data, including: an information recognition module, an address information processing module, an address two-way matching module, and an abnormal address information processing and decision-making module;

[0006] The information recognition module is used to collect an image containing information on the student financial aid application form to obtain an input image; and based on the OCR technology, perform text recognition on the input image to output a recognition result of the student address information;

[0007] The address information processing module is used to split and parse each student's address information according to different levels based on the student address information recognition results and the historical student address database to obtain the corresponding standard address information; and to determine the level status of all students' standard address information, establish an address level fixed collocation matching detection model based on the determination results, and generate corresponding processing signals based on the address level fixed collocation matching detection;

[0008] The address bidirectional matching module is used to perform statistical analysis on the abnormal address information under all address abnormal signals, identify the corresponding top-level address information and bottom-level address information, and generate forward search results and reverse search results by combining the preset forward deduction rules and reverse deduction rules;

[0009] The abnormal address information processing and decision module is used to judge the consistency of the forward search results and reverse search results corresponding to each abnormal address information, and implement the corresponding abnormal address information processing strategy according to the judgment result; the abnormal address processing strategy includes automatic correction strategy and abnormal prompt modification strategy.

[0010] As a further solution of the present invention: the logic for obtaining the student standard address information is: based on the student address information recognition result, and in combination with the historical student address database, each student address information is split and analyzed according to different levels to obtain the corresponding student standard address information that stores multiple levels of standard fixed combinations.

[0011] As a further solution of the present invention: judging the hierarchical status of all students' standard address information and establishing an address hierarchical fixed collocation matching detection model is specifically as follows:

[0012] S01: Define the corresponding address level in each student's standard address information to obtain the address level sequence Xs=[L1, L2, ..., L n ]; where s represents the standard address information of each student, n is the number of levels and 1≤n≤5, i is the identifier of different levels, L i Indicates different levels;

[0013] Each address level sequence is then precisely compared with the fixed collocation of the corresponding student address levels in the historical student address database to obtain the L i Matching state function Among them, U1 is the level L i Exists and the name matches the fixed collocation in the historical student address database, U2 is level L i Missing or incorrect name; the status value contains 1 and 0, indicating level L respectively. i The corresponding matching status is normal and abnormal;

[0014] S02: Based on the above state function, the matching state function of the entire address hierarchy sequence is obtained. Among them, MS(Xs) represents the matching status of the entire address level sequence, and n represents the number of levels;

[0015] S03: Based on the matching state function, establish an address-level fixed collocation matching detection model Among them, U1 and U2 represent complete address level fixed collocation matching and incomplete level fixed collocation matching, respectively;

[0016] When AD=U3, the normal address signal is generated and the application form approval stage is directly entered;

[0017] When AD=U4, an address abnormality signal is generated and the corresponding student standard address information is marked as abnormal address information.

[0018] As a further solution of the present invention: the specific process of generating the forward search result and the reverse search result is: select an abnormal address information as the target abnormal address information, follow the preset forward deduction rule to obtain the existing address sequence Xy=[L1,L m ]; where y represents each abnormal address information. From the historical student address database, all addresses that meet the above existing address hierarchy sequence are screened and recorded as candidate address sequence Xh. h represents the number of candidate addresses. Based on the existing address sequence Xy=[L1,L m ] The middle level address information of the abnormal address level sequence is represented as Xyz=L2,...,L m-1 ]; Similarly, the middle-level address information Xhz of the candidate address sequence is obtained;

[0019] For each candidate address sequence, the string similarity algorithm is used to calculate the similarity between the intermediate level address information Xhz of the candidate address sequence and the intermediate level address information Xyz of the abnormal address level sequence, that is, Among them, LD is the edit distance between the two strings, len(Xhz) and len(Xyz) are the lengths of the candidate address sequence and the existing address hierarchy sequence respectively. The similarity of each candidate address is compared and the address with the highest similarity is used as the target abnormal address information forward search result Z g1 , record its similarity value S g1 ; Where g is the address identifier of the target abnormal address information;

[0020] The above forward search result derivation process is similarly implemented for all abnormal address information to obtain the forward search result Z corresponding to each abnormal address information. y1 And its corresponding similarity S y1 ;

[0021] Follow the preset reverse deduction rules to obtain the existing address hierarchy sequence containing the lowest level address and the highest level address Consistent with the forward search result acquisition logic, all addresses that meet the above existing address hierarchy sequence are screened from the historical student address database and recorded as candidate address sequence Xf;

[0022] For each candidate address sequence, use the string similarity algorithm to calculate the intermediate level address information Xfz of the candidate address sequence and the intermediate level address information of the abnormal address level sequence The address with the highest similarity is used as the reverse lookup result Z g2 , record its similarity value S g2 ;

[0023] The above reverse lookup result derivation process is implemented for all abnormal address information to obtain the reverse lookup result Z of the target abnormal address information corresponding to each abnormal address information. y2 And its corresponding similarity S y2 .

[0024] As a further solution of the present invention, the process of determining the consistency between the forward search result and the reverse search result corresponding to each abnormal address information is as follows: all forward search results Z generated in the address bidirectional matching module are matched. y1 、Reverse search result Z y2 And its corresponding similarity S y1 and S y2 Collect statistics and form a two-way matching set corresponding to each abnormal address information [(Z y1 , S y1 ),(Z y2 , S y2 )], and perform internal comparison on the similarity of the two-way matching set of each abnormal address information in turn. Specifically:

[0025] When S y1 =S y2 , the forward search results and the reverse search results are consistent; otherwise they are inconsistent.

[0026] As a further solution of the present invention: the corresponding abnormal address information processing strategies are implemented according to the judgment result and the preset similarity threshold q1:

[0027] When S y1 ≠S y2 When an error occurs, the address logic error signal is directly generated to execute the abnormal prompt modification strategy;

[0028] When S y1 =S y2 , S y1 orS y2 ≤q1, then execute the automatic correction strategy;

[0029] When S y1 =S y2 , S y1 orS y2 >q1, enter the special judgment unit for execution strategy.

[0030] As a further solution of the present invention: the corresponding abnormal address information processing strategies are implemented according to the judgment result and the preset similarity threshold q1:

[0031] When S y1 ≠S y2 When an error occurs, the address logic error signal is directly generated to execute the abnormal prompt modification strategy;

[0032] When S y1 =S y2 , S y1 orS y2 ≤q1, then execute the automatic correction strategy;

[0033] When S y1 =S y2 , S y1 orS y2 >q1, enter the special judgment unit for execution strategy.

[0034] Compared with the existing technology, the advantages of the present invention are:

[0035] 1. The present invention identifies the address information provided by students and decomposes and analyzes it in combination with the historical student address database to obtain standard address information. Subsequently, this information is compared with the standard address fixed collocation in the historical database, and an address hierarchical fixed collocation matching detection model is constructed. The address hierarchical fixed collocation matching detection model is traversed and the matching status of all hierarchies is detected to generate complete address signals and abnormal address signals. The abnormal address information under various address abnormality signals is statistically analyzed to identify the corresponding top-level address information and bottom-level address information. In combination with preset forward deduction rules and reverse deduction rules, forward and reverse search results are generated. Based on the forward and reverse search results and their similarity values, a comprehensive matching analysis is performed on all abnormal address information. This can not only effectively identify the integrity and accuracy of the address information, provide an accurate basis for the subsequent funding application review, but also improve the accuracy and efficiency of address information review and correction;

[0036] 2. The present invention collects the forward search results and reverse search results corresponding to each abnormal address information to make consistency judgments, thereby realizing automatic correction of abnormal address information or abnormal prompt modification; for some special cases, a comprehensive judgment and scoring mechanism is used to further judge whether to execute the automatic correction strategy or the abnormal prompt modification strategy. The specific basis is to calculate the overall similarity and the number of characters of the highest character, and combine the preset weight and balance coefficient to obtain a comprehensive judgment score. Finally, a strategy selection decision is made based on the comprehensive judgment score and the preset threshold, effectively improving the accuracy of address information and the review efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1This is the overall structural diagram of the intelligent student aid application review and approval system based on big data proposed by the present invention;

[0038] Figure 2 This is an execution flow chart of the special judgment unit of the execution strategy of the big data-based intelligent student aid application review and approval system proposed by the present invention. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0040] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside", "outside", etc., indicating directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0041] Reference Figure 1-Figure 2 The intelligent student aid application review and approval system based on big data includes: information identification module, address information processing module, address two-way matching module, abnormal address information processing and decision-making module;

[0042] The information recognition module is used to capture an image containing information on a student financial aid application form to obtain an input image; and based on OCR technology, perform text recognition on the input image to output a student address information recognition result; including:

[0043] Based on an image containing a student financial aid application form received by the system, the image containing the student financial aid application form is used as an input image for subsequent text recognition; wherein the student financial aid application form includes basic information of the student, family information, type of financial aid applied for, and reason;

[0044] Based on OCR technology, the input image is subjected to text recognition, and the student address information recognition result is output; wherein, the OCR technology is based on the PaddleOCR model framework; specifically, the process of using the PaddleOCR model framework to perform text recognition on the input image is as follows:

[0045] Detecting text regions in the input image using a preset image analysis method to obtain regions containing text; wherein the preset image analysis method may be, for example, a Faster R-CNN algorithm or a YOLO algorithm; it should be noted that in the process of obtaining regions containing text, the input image is analyzed and processed based on color, shape, or texture features using the preset image analysis method to determine regions in the input image that may contain text;

[0046] Using the pre-trained address recognition model, address information features are extracted from several single characters or character groups to convert address feature information such as the shape and outline of the characters or character groups into computer-recognizable data values, thereby obtaining the student address information recognition results.

[0047] It should be noted that the above-mentioned OCR technology is optical character recognition technology, which refers to the process in which an electronic device checks the characters printed on paper, determines their shapes by detecting dark and light patterns, and then uses character recognition methods to translate the shapes into computer text. This is a mature technology in the existing technology. In this embodiment, the specific process of recognizing student address information is not described in detail.

[0048] The address information processing module is used to split and parse each student's address information according to different levels based on the student address information recognition results and the historical student address database to obtain the corresponding standard address information; and to determine the level status of all students' standard address information, establish an address level fixed collocation matching detection model based on the determination results, and generate corresponding processing signals based on the address level fixed collocation matching detection model;

[0049] Based on the student address information recognition result, and in combination with the historical student address database, each student address information is split and parsed according to different levels to obtain the corresponding student standard address information; the historical student address database is a database pre-stored in the system that contains various basic information of students and stores standard address fixed collocations, that is, it contains accurate and standardized address level fixed collocation examples of the student's corresponding address, such as "Guangdong Province -> Guangzhou City -> Tianhe District -> Shipai Street -> a certain house number" such a complete and correct address information fixed collocation;

[0050] An example of a fixed collocation of address levels is that the student standard address information is stored with multiple hierarchical fields, and each of the student standard address information contains at most five hierarchical fields; the first hierarchical field is province / municipality / autonomous region, the second hierarchical field is prefecture-level city / district, the third hierarchical field is county-level city / county, the fourth hierarchical field is county / district, and the fifth hierarchical field is township and street office or village committee / residential committee / number, which can be abbreviated as province, city, county / district, town or village;

[0051] The hierarchical status of all students' standard address information is judged, and an address hierarchical fixed collocation matching detection model is established. Specifically:

[0052] S01: Define the corresponding address level in each student's standard address information to obtain the address level sequence Xs=[L1, L2, ..., L n ]; where s represents the standard address information of each student, n is the number of levels and 1≤n≤5, i is the identifier of different levels, L i Indicates different levels;

[0053] Each address level sequence is then precisely compared with the fixed collocation of the corresponding student address levels in the historical student address database to obtain the L i Matching state function Among them, U1 is the level L i Exists and the name matches the fixed collocation in the historical student address database, U2 is level L i Missing or incorrect name; the status value contains 1 and 0, indicating level L respectively. i The corresponding matching status is normal and abnormal;

[0054] S02: Based on the above state function, the matching state function of the entire address hierarchy sequence is obtained. Among them, MS(Xs) represents the matching status of the entire address level sequence, and n represents the number of levels;

[0055] S03: Based on the matching state function, establish an address-level fixed collocation matching detection model Among them, U1 and U2 represent complete address level fixed collocation matching and incomplete level fixed collocation matching, respectively;

[0056] When AD=U3, it means that all levels exist and the names are correct, the addresses are complete, and a normal address signal is generated to directly enter the application form approval stage;

[0057] When AD=U4, it means that at least one level is missing or the name is wrong, the address is incomplete, and an address abnormality signal is generated and the corresponding student standard address information is marked as abnormal address information, entering the abnormal address information processing and analysis stage;

[0058] It should be noted that when AD=U3, that is, there are five levels of hierarchical fields (province, city, county, town, village) at the same time, then the address information is complete. When AD=U4, there is at least one level field (city, county, town, village) missing.

[0059] The address bidirectional matching module is used to perform statistical analysis on the abnormal address information under all address abnormal signals, identify the corresponding top-level address information and bottom-level address information, and generate forward search results and reverse search results by combining the preset forward deduction rules and reverse deduction rules;

[0060] Get the abnormal address level sequence Xy corresponding to each abnormal address information = [L1, L2, ..., L m ]; where y represents each abnormal address information, m is the number of levels and 1≤m≤4, j is the identifier of different levels, L j Indicates different levels;

[0061] For the abnormal address level sequence Xy=[L1,L2,...,L m ] to traverse, define the first element L1 (highest level, such as province / municipality) of the sequence as the top-level address information; define the last element Lm (lowest level, such as street / house number) of the sequence as the bottom-level address information;

[0062] The specific process of generating forward search results and reverse search results is as follows: first select an abnormal address information as the target abnormal address information, follow the preset forward deduction rule, that is, start from the top level L1, for each level L j , traverse and check L j Does the subordinate include L j+1 , when included, continue to deduce down to the next level until you get the lowest level L m , get the existing address sequence Xy=[L1,L m ]; where y represents each abnormal address information. From the historical student address database, all addresses that meet the above existing address hierarchy sequence are screened and recorded as candidate address sequences. Xhh represents the number of candidate addresses. Based on the existing address sequence Xy=[L1,L m ] The middle level address information of the abnormal address level sequence is represented as Xyz=[L2,...,L m-1 ]; Similarly, the middle-level address information Xhz of the candidate address sequence is obtained;

[0063] For each candidate address sequence, the string similarity algorithm is used to calculate the similarity between the intermediate level address information Xhz of the candidate address sequence and the intermediate level address information Xyz of the abnormal address level sequence, that is, Among them, LD is the edit distance between the two strings, which is directly calculated by the system. len(Xhz) and len(Xyz) are the lengths of the candidate address sequence and the existing address hierarchy sequence respectively. The similarity of each candidate address is compared and the address with the highest similarity is used as the forward search result Z. g1 , record its similarity value S g1; Where g is the address identifier of the target abnormal address information;

[0064] The above forward search result derivation process is similarly implemented for all abnormal address information to obtain the forward search result Z corresponding to each abnormal address information. y1 And its corresponding similarity S y1 ;

[0065] Reverse lookup result Z y2 :Follow the preset reverse deduction rule, starting from the lowest level Lm, for each level L j , traverse and check L j Does the subordinate include L j-1 If it is included, continue to deduce down to the next level until the top level L1 is obtained, and get the existing address level sequence containing the lowest level address and the highest level address Consistent with the forward search result acquisition logic, all addresses that meet the above existing address hierarchy sequence are screened from the historical student address database and recorded as candidate address sequence Xf;

[0066] For each candidate address sequence, use the string similarity algorithm to calculate the intermediate level address information Xfz of the candidate address sequence and the intermediate level address information of the abnormal address level sequence The similarity S y2 , compare the similarity of each candidate address and take the address with the highest similarity as the reverse lookup result Z of the target abnormal address information y1 , record its similarity value S y2 ;

[0067] The above forward search result and reverse search result derivation process is implemented for all abnormal address information in the same way to obtain the forward search result Z corresponding to each abnormal address information. y1 、Reverse search result Z y2 And its corresponding similarity S y1 and S y2 .

[0068] Exemplarily, the logic for obtaining the forward search results and the reverse search results in this example takes the forward search results as an example: when the forward existing address hierarchy sequence Xy = [L1, L2, L3]; where the top level L1 = Guangdong Province, the bottom level L3 = Nanshan Subdistrict, and L2 = Shenzhen City, all addresses that meet the above existing address hierarchy sequence are screened from the historical student address database and recorded as candidate address sequences Xh. For example, when the screened candidate address sequences are: Guangdong Province - Shenzhen City - Nanshan District - Nanshan Subdistrict, Guangdong Province - Guangzhou City - Tianhe District - Nanshan Subdistrict; calculate the similarity S between the other address information in the above candidate addresses except the top level address information and the bottom level address information and the middle level (Shenzhen City) in the existing candidate address information. 11 and S21 , where S 11 >S 21 That is, the address with the highest similarity is Guangdong Province - Shenzhen City - Nanshan District - Nanshan Street, which is recorded as the forward search result Z 11 .

[0069] The abnormal address information processing and decision module is used to judge the consistency of the forward search results and reverse search results corresponding to each abnormal address information, and implement the corresponding abnormal address information processing strategy according to the judgment results;

[0070] Specifically: the abnormal address processing strategy includes an automatic correction strategy and an abnormal prompt modification strategy;

[0071] All forward search results Z generated in the address bidirectional matching module y1 、Reverse search result Z y2 And its corresponding similarity S y1 and S y2 Collect statistics and form a two-way matching set corresponding to each abnormal address information [(Z y1 , S y1 ),(Z y2 , S y2 )], perform internal comparison on the similarity of the two-way matching set of each abnormal address information in turn and make strategy selection based on the preset similarity threshold q1. The specific value is set by researchers according to actual needs; specifically:

[0072] When S y1 ≠S y2 When the address logic error signal is directly generated, the abnormal prompt modification strategy is executed, that is, the student financial aid application form described in the abnormal address information is fed back to the user interface, allowing the user to check and modify it;

[0073] When S y1 =S y2 ,S y1 orS y2 ≤q1, the abnormal address information corresponding to the bidirectional matching set is automatically corrected based on the selected candidate address sequence; characterize Z y1 and Z y2 The candidate address level is completely matched and meets the similarity threshold. The system executes the automatic correction strategy to automatically correct the address directly based on the candidate address.

[0074] When S y1 =S y2 ,S y1 orS y2 >q1, further judgment is needed on whether to execute the automatic correction strategy or the abnormal prompt modification strategy;

[0075] In the abnormal address information processing and decision-making module, an execution policy special judgment unit is further included for when S y1 = S y2 , S y1 or S y2 > q1, it is necessary to further judge whether to execute the automatic correction policy or the abnormal prompt modification policy;

[0076] When S y1 [[ID=,13]]= S y2 , S y1 or S y2 > q1, this unit is triggered. Specifically, for each two-way matching set [(Z y1 , S y1 ), (Z y2 , S y2 )] corresponding to each abnormal address information, when S y1 = S y2 , S y1 or S y2 > q1, calculate the similarity values S ym of each level in the corresponding abnormal address information in sequence, and calculate the average value of the similarities of each level to obtain the overall similarity At the same time, record the number of characters of the level with the most characters in each abnormal address information, denoted as the number of characters N y ;

[0077] According to the obtained overall similarity and the number of characters N y of the highest character, calculate the comprehensive judgment score ω1 and ω2 are the weights of the overall similarity and the number of characters N y of the highest character respectively. Their specific values are set by researchers according to actual needs, and q char is the balance coefficient;

[0078] Based on the comprehensive judgment score and the preset threshold q2, construct the comprehensive judgment rule for policy selection. Specifically:

[0079] When P≥q2, execute the automatic correction policy, and the system modifies the address according to the candidate address;

[0080] When P < q2, execute the abnormal prompt modification policy, and color the address to prompt the user of the address error; indicating that the average similarity is insufficient and the highest character is abnormal.

[0081] In this example, an image containing student aid application form information is collected through the information recognition module, and OCR technology is used to recognize the text in the input image, extract the student address information, and output the recognition result; the address information processing module splits and analyzes the address information based on the student address information recognition result and the historical student address database to form standard address information, and judges its hierarchical status, establishes an address hierarchy fixed matching detection model, and generates a processing signal; the address two-way matching module statistically analyzes all address abnormality signals, identifies the top-level and bottom-level addresses of the abnormal address information, and generates forward search results and reverse search results by combining the preset forward deduction rules and reverse deduction rules; the abnormal address information processing and decision module compares the consistency of the forward and reverse search results of each abnormal address information, applies the corresponding processing strategy according to the judgment result, and further decides whether to automatically correct or prompt modification through the special situation judgment unit, thereby realizing efficient and accurate operation of intelligent review and approval.

[0082] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. Intelligent student aid application review and approval system based on big data, It is characterized in that it includes: An information recognition module is used to collect an image containing the information of the student financial assistance application form to obtain an input image; and based on OCR technology, perform character recognition on the input image and output the recognition result of the student address information. The address information processing module is used to split and analyze each student address information according to different levels in combination with the historical student address database based on the recognition result of the student address information to obtain the corresponding standard address information; and judge the level status of the standard address information of all students, establish an address level fixed collocation matching detection model according to the judgment result, and generate corresponding processing signals according to the address level fixed collocation matching detection. The address two-way matching module is used to statistically analyze the abnormal address information under all address abnormal signals, identify its corresponding top-level address information and bottom-level address information, and generate a forward search result and a reverse search result in combination with the preset forward derivation rule and reverse derivation rule. The abnormal address information processing and decision-making module is used to judge the consistency of the forward search result and the reverse search result corresponding to each abnormal address information, and implement corresponding abnormal address information processing strategies according to the judgment result; the abnormal address processing strategy includes an automatic correction strategy and an abnormal prompt modification strategy.

2. The intelligent student aid application review and approval system based on big data according to claim 1 is characterized by: The student standard address information is: based on the recognition result of the student address information, and splitting and analyzing each student address information according to different levels in combination with the historical student address database to obtain the student standard address information storing multiple levels of standard fixed collocations.

3. The intelligent student aid application review and approval system based on big data according to claim 1 is characterized in that: It includes: Judging the level status of all student standard address information and establishing an address level fixed collocation matching detection model specifically: S01: Define the corresponding address level in each student's standard address information to obtain the address level sequence Xs=[L1, L2, ..., L n ]; where s represents the standard address information of each student, n is the number of levels and 1≤n≤5, i is the identifier of different levels, L i Indicates different levels; And then accurately compare each address level sequence with the corresponding student address level fixed collocation in the historical student address database to obtain each level L i Matching state function Among them, U1 is the level L i Exists and the name matches the fixed collocation in the historical student address database, U2 is level L i Missing or incorrect name; the status value contains 1 and 0, indicating level L respectively. i The corresponding matching status is normal and abnormal; S02: Based on the above state function, the matching state function of the entire address hierarchy sequence is obtained. Among them, MS(Xs) represents the matching status of the entire address level sequence, and n represents the number of levels; S03: Based on the matching state function, establish an address-level fixed collocation matching detection model Among them, U1 and U2 represent complete address level fixed collocation matching and incomplete level fixed collocation matching, respectively; When AD = U3, generate an address normal signal and directly enter the application form approval stage; When AD = U4, generate an address abnormal signal and mark the corresponding student standard address information as abnormal address information.

4. The intelligent student aid application review and approval system based on big data according to claim 1 is characterized in that: It includes: The specific process of generating forward search results and reverse search results is as follows: select an abnormal address information as the target abnormal address information, follow the preset forward deduction rule to obtain the existing address sequence Xy=[L1,L m ]; where y represents each abnormal address information. From the historical student address database, all addresses that meet the above existing address hierarchy sequence are screened and recorded as candidate address sequence Xh. h represents the number of candidate addresses. Based on the existing address sequence Xy=[L1,L m ] The middle level address information of the abnormal address level sequence is represented as Xyz=[L2,...,L m-1 ]; Similarly, the middle-level address information Xhz of the candidate address sequence is obtained; For each candidate address sequence, the string similarity algorithm is used to calculate the similarity between the intermediate level address information Xhz of the candidate address sequence and the intermediate level address information Xyz of the abnormal address level sequence, that is, Among them, LD is the edit distance between the two strings, len(Xhz) and len(Xyz) are the lengths of the candidate address sequence and the existing address hierarchy sequence respectively. The similarity of each candidate address is compared and the address with the highest similarity is used as the target abnormal address information forward search result Z g1 , record its similarity value S g1 ; Where g is the address identifier of the target abnormal address information; The above forward search result derivation process is similarly implemented for all abnormal address information to obtain the forward search result Z corresponding to each abnormal address information. y1 And its corresponding similarity S y1 ; Follow the preset reverse deduction rules to obtain the existing address hierarchy sequence containing the lowest level address and the highest level address Consistent with the forward search result acquisition logic, all addresses that meet the above existing address hierarchy sequence are screened from the historical student address database and recorded as candidate address sequence Xf; For each candidate address sequence, use the string similarity algorithm to calculate the intermediate level address information Xfz of the candidate address sequence and the intermediate level address information of the abnormal address level sequence The address with the highest similarity is used as the reverse lookup result Z g2 , record its similarity value S g2 ; The above reverse lookup result derivation process is implemented for all abnormal address information to obtain the reverse lookup result Z of the target abnormal address information corresponding to each abnormal address information. y2 And its corresponding similarity S y2 .

5. The intelligent student aid application review and approval system based on big data according to claim 1 is characterized in that: It includes: The process of judging the consistency of the forward search results and reverse search results corresponding to each abnormal address information is as follows: all forward search results Z generated in the address bidirectional matching module are matched. y1 、Reverse search result Z y2 And its corresponding similarity S y1 and S y2 Collect statistics and form a two-way matching set corresponding to each abnormal address information [(Z y1 , S y1 ),(Z y2 , S y2 )], and perform internal comparison on the similarity of the two-way matching set of each abnormal address information in turn. Specifically: When S y1 =S y2 , the forward search results are consistent with the reverse search results; Otherwise, they are inconsistent.

6. The intelligent student aid application review and approval system based on big data according to claim 5 is characterized in that: It includes: implementing corresponding abnormal address information processing strategies according to the judgment result and the preset similarity threshold q1 respectively: When S y1 ≠S y2 When an error occurs, the address logic error signal is directly generated to execute the abnormal prompt modification strategy; When S y1 =S y2 , S y1 orS y2 ≤q1, then execute the automatic correction strategy; When S y1 =S y2 , S y1 orS y2 >q1, does not meet any policy execution conditions, and enters the special judgment unit for executing the policy.

7. The big data-based intelligent student aid application review and approval system according to claim 6 is characterized by: The abnormal address information processing and decision module also includes an execution strategy special judgment unit for when S y1 =S y2 ,S y1 orS y2 >q1, further judgment is needed on the choice of execution strategy; When S y1 =S y2 ,S y1 orS y2 >q1, trigger the unit, for each abnormal address information corresponding to the bidirectional matching set [(Z y1 , S y1 ),(Z y2 , S y2 )], when S y1 =S y2 ,S y1 orS y2 When >q1, calculate the similarity value S of each level in the corresponding abnormal address information in turn. ym , and calculate the average value of the similarity at each level to get the overall similarity At the same time, the number of characters in the level with the most characters in each abnormal address information is recorded as the number of characters N of the highest characters. y ; According to the overall similarity and the number of characters N of the highest character y Calculate the comprehensive judgment score ω1 and ω2 are the overall similarity and the number of characters N of the highest character y The weight, q char is the balance coefficient; Based on the comprehensive judgment score and the preset threshold q2, constructing a strategy selection comprehensive judgment rule specifically: When P ≥ q2, execute the automatic correction strategy; When P < q2, execute the abnormal prompt modification strategy to prompt the user that the address is incorrect.

Images