Inspection and supervision clue knowledge graph mining and analysis method for economic and financial supervision

Abstract

This invention relates to the field of data relationship mining technology, specifically to a knowledge graph mining and analysis method for inspection and supervision clues in economic and financial supervision. It obtains corresponding results by acquiring inspection clues and data such as funds and equity, constructs a knowledge graph network containing entity and account associations, compares the control hierarchy of the subject with the order of fund flow to filter for unidirectional paths, accurately identifies consistent fund transfers and closed interval segments, and finally arranges relevant subjects hierarchically and displays them with connections to generate knowledge graph mining and analysis results. This invention constructs clue segments based on the co-occurrence of unit names and times, organizes transaction control relationships by account subject, achieves precise alignment of clues and data with a unified identifier, generates consistent fund mappings for clues by combining time filtering, constructs unidirectional associations by comparing control hierarchy and flow order, extracts unidirectional closed-loop paths of consistent intervals, and displays cyclical trajectories to enhance the ability to identify and interpret associations.

Description

Technical Field

[0001] This invention relates to the field of data relationship mining technology, and in particular to a method for mining and analyzing knowledge graphs of clues from inspection and supervision for economic and financial oversight. Background Technology

[0002] The field of data relationship mining technology encompasses data acquisition, data modeling, relationship representation, and association analysis. This technology focuses on the organization and expression of structured and unstructured data. By performing entity recognition, attribute extraction, relationship construction, and graph structure representation on multi-source data, it achieves a systematic characterization of the relationships between complex data. Its overall technical system covers rule-based entity matching methods, statistical feature-based relationship extraction methods, and graph-oriented path analysis and pattern recognition methods. Combined with data storage index construction and multi-dimensional relationship query mechanisms, it provides a unified description and analysis of cross-domain and cross-level data relationships.

[0003] Among them, the knowledge graph mining and analysis method for inspection and supervision clues in economic and financial supervision refers to a processing method that uses a unified modeling and relationship construction method for inspection and supervision clue data in economic and financial supervision scenarios, such as audit records, fund flows to enterprises, related personnel information, and project approval records. The technical matters involved include entity labeling of institution names, personnel identities, fund amounts, and time nodes in inspection and supervision text records; labeling and aligning the same entity in different data sources; constructing relationship triples for fund transfers, employment relationships, and project participation relationships between entities; serializing the relationships based on time order; combining and parsing multi-hop relationships through preset relationship path rules; and filtering and labeling relationship paths in the knowledge graph based on defined abnormal relationship patterns.

[0004] Existing technologies primarily employ relation extraction and path combination as analytical methods, focusing on rule matching and structural expression during processing. However, they lack a unified screening mechanism around time intervals, leading to a risk of misalignment between clues and financial behavior in the time dimension. In scenarios involving multiple entities and accounts, it is difficult to form a continuous sequence of fund transfer paths, and the presentation of relationships tends to remain at the level of discrete fragments. Furthermore, the lack of sequential comparison constraints between control relationships and fund flows makes it difficult to uniformly determine the direction of upstream and downstream transmission. In cases involving multiple round trips, it is impossible to effectively identify closed intervals or cyclical paths, thereby affecting the completeness and reliability of association judgments. Summary of the Invention

[0005] To address the technical problems existing in current technologies, this invention provides a method for mining and analyzing knowledge graphs of inspection and supervision clues for economic and financial oversight. The technical solution is as follows: A knowledge graph mining and analysis method for inspection and supervision clues aimed at economic and financial supervision includes the following steps: S1: Obtain clue text, fund transaction flow and corporate equity and appointment information, segment the clue text into clue fragments, divide continuous transaction fragments and control fragments, identify the subject name and establish the association between the subject and the account and relationship identifier, combine time period to filter and match, and generate the corresponding results of the funds of the inspection and supervision clues. S2: Based on the account and entity identifiers in the fund correspondence results of the inspection and supervision clues, construct the enterprise entity, natural person entity and fund account graph points, establish the connection between investment relationship, appointment relationship and fund flow, associate the clue fragment description content, and generate an economic and financial supervision knowledge graph network; S3: Based on the connections of investment, appointment and fund flow in the economic and financial supervision knowledge graph network, construct the main control relationship network and fund destination path, compare the hierarchical number and sequence number, select the main graph points and connections with the same direction, and generate the control level fund sequence analysis results. S4: Based on the fund destination path and sequential number in the fund sequence analysis results of the control level, connect them to form an ordered account sequence, identify the same account identifier and determine the interval segment, extract the segment with the same first and last account identifier and classify the type, and generate the inspection and patrol cycle round-trip path results.

[0006] As a further aspect of the present invention, the results of the correspondence between the funds and the inspection clues include time range identifiers, corresponding groups of main accounts, clue association units, data matching markers, and effective combination sets; the economic and financial supervision knowledge graph network includes enterprise main nodes, natural person main nodes, fund account nodes, relationship connection sets, and semantic description information; the results of the control level fund sequence analysis include control level sequence, fund path sequence, direction consistency markers, level correspondence relationships, and effective path sets; and the results of the inspection and inspection cyclic round-trip path include fund round-trip segments, fund closure segments, path interval identifiers, path classification markers, and path-related main entities.

[0007] As a further aspect of the present invention, the step of obtaining S1 is as follows: S11: Obtain the text of inspection and supervision clues from the inspection and supervision office terminal, as well as the fund transaction flow, corporate equity and appointment information from the data collection server of the regulatory data center. Call the period, semicolon and comma positions in the text of inspection and supervision clues to divide the sentences into segments, extract the unit name, personnel name and time period sentence by sentence, and merge them according to the number of times the same sentence appears to obtain co-occurrence segment groups. S12: Based on the co-occurrence fragment group, call the account identifier, transaction time, entity name, and document number in the fund transaction flow, and check the entity name, document number, and entity relationship identifier in the enterprise equity and appointment information item by item. Merge and mark the items with the same name and the items with the same document number, and write the account identifier and entity relationship identifier into the same entity identifier set to obtain the associated list. S13: Based on the associated directory, call the account identifier and transaction time in the continuous transaction segment, call the subject relationship identifier and subject identifier in the control segment, perform interval comparison for the start and end time of the time period in the clue segment, select the combination of account identifier and subject identifier that falls in the same time interval and appears synchronously, and generate the corresponding result of the inspection and supervision clue funds.

[0008] As a further aspect of the present invention, the step of obtaining S2 is as follows: S21: Based on the account identifier and subject identifier in the fund correspondence results of the inspection and investigation clues, call the subject identifier entries and account identifier entries in the map storage device for classification and judgment, write the subject identifier into the node sequence according to the enterprise registration mark and the natural person registration mark, write the account identifier into the account sequence according to the payment mark, and assign map point numbers to the node sequence and the account sequence in sequence to establish map point catalog; S22: Based on the above point catalog, retrieve the entries for investor name, invested entity name, employing entity name, and employing status from the enterprise equity and employment information, check the corresponding positions of point numbers and entity identifiers item by item, connect the enterprise entity points with the investment direction line, connect the natural person entity points with the enterprise entity points with the employment direction line, retrieve the payment account identifier and receiving account identifier from the fund transaction flow and write them into the flow direction line between the account points to obtain the relationship network column; S23: Based on the aforementioned relationship network column, call the inspection and supervision clue text to form the description content in the clue fragment, and attach it to the enterprise main body map point or natural person main body map point according to the main body identifier. Perform adjacency verification on the enterprise main body map point, natural person main body map point, fund account map point and the start and end numbers of each connecting line, write the consecutive numbered items into the same map network column, write the description content into the corresponding description position, and generate an economic and financial supervision knowledge graph network.

[0009] As a further aspect of the present invention, the step of obtaining S3 is as follows: S31: Construct a subject control relationship network based on the directed connections of investment relationships and appointment relationships in the economic and financial supervision knowledge graph network. Call up enterprise subject graph points, natural person subject graph points, investment relationship start and end numbers, appointment relationship start and end numbers, filter out graph points with upstream control source numbers, register the next hop graph point number layer by layer along the external connection direction of the remaining graph points, and write it into the hierarchical position according to the transmission order to establish a control layer sequence; S32: Based on the control layer sequence, call the payment account identifier, receiving account identifier, transaction time record and fund account point number in the directed connection of fund flow, perform time sorting for the external account points corresponding to the same payment account identifier, connect the previous receiving account point and the subsequent payment account point in sequence, and write the order position according to the connection order, call the enterprise main point number, natural person main point number and hierarchical position to compare with the fund account point number and order position to obtain the same direction list; S33: Based on the aforementioned list of parallel paths, call the enterprise entity map point number, natural person entity map point number, investment relationship connection number, employment relationship connection number, capital account map point number, hierarchical position, and sequential position. For items in the same column as the transmission direction and the sending direction, perform the attachment, write the hierarchical position and sequential position into the same presentation sequence, write the enterprise entity map point, natural person entity map point and corresponding connection into the associated position, and generate the control level capital sequence analysis result.

[0010] As a further aspect of the present invention, the filtering out of points with upstream control source numbers specifically involves: The corresponding positions of the enterprise entity map points and the natural person entity map points in the start and end numbers of the investment relationship and the start and end numbers of the employment relationship are checked. Map points that appear as termination numbers in the start and end numbers of the investment relationship or the start and end numbers of the employment relationship and have corresponding start numbers are determined to be map points with upstream control source numbers and are removed from the traversal range. The specific steps for registering the next-hop point number layer by layer along the external connection direction of the remaining point are as follows: Based on the connection relationship between the start and end numbers of the investment relationship and the start and end numbers of the appointment relationship, a layer-by-layer traversal is performed, and each termination number is written into the corresponding level position in the connection order. The map point numbers that appear repeatedly in the same level position are merged and recorded.

[0011] As a further aspect of the present invention, the sorting of execution time for external account graph points corresponding to the same payment account identifier specifically includes: The transaction time records are arranged sequentially, and the connection between adjacent records after sorting is judged. Account points whose receiving account identifiers are consistent with the paying account identifiers in subsequent records are established as continuous connections and written into the order position.

[0012] As a further aspect of the present invention, the step of obtaining S4 is as follows: S41: Based on the fund destination path and corresponding sequence number obtained from the fund sequence analysis results of the control level, call the starting fund account map point, subsequent fund account map points, and sequence number, perform a connection judgment for adjacent sequence numbers, write the fund account map points with continuous sequence and consistent connection direction into the same arrangement sequence, and record the corresponding position number of each fund account map point to establish an account arrangement list. S42: Based on the account ranking list, call the account identifier and position number corresponding to each fund account point, perform position comparison for fund account points with the same account identifier, write the range between the previous position number and the next position number into the corresponding interval segment, perform consistency judgment on the first account identifier and the last account identifier, extract the consistent interval segments and register them separately according to the round-trip direction and the closed direction to obtain the segment list; S43: Based on the fragment list, call up the enterprise main body map points, natural person main body map points, level number, and interval fragment position number, perform corresponding verification for the map point appearance position in each interval fragment, write the enterprise main body map points or natural person main body map points into the corresponding interval fragment association position, write the level number into the same fragment label position, and combine the fund round trip fragments and closed fragments to generate the inspection and patrol cycle round trip path result.

[0013] As a further aspect of the present invention, the method further includes: S5: Based on the fund round-trip segments, closed segments, fund destination paths and main map point hierarchical numbers in the inspection and patrol cycle round-trip path results, select relevant map points and arrange them in layers, connect the fund destination lines, display the corresponding fund destination paths respectively, and generate knowledge graph mining analysis results; The knowledge graph mining and analysis results include a hierarchical display structure, path display units, key node sets, relationship connection sets, and analysis result views.

[0014] As a further aspect of the present invention, the step of obtaining S5 is as follows: S51: Based on the fund transfer segments, closed segments, fund destination paths, and the association between enterprise entity map points, natural person entity map points, and hierarchical numbers in the results of the patrol and inspection cycle, call the segment identifier, map point number, hierarchical number, and path number in the map display device, perform map point aggregation judgment for fund transfer segments and closed segments, and write enterprise entity map points, natural person entity map points, and fund account map points into the same display selection sequence to establish a map point selection sequence; S52: Based on the selected point sequence, call the enterprise main point number, natural person main point number, level number, and fund account point number. Sort the level number by position, write the point with the earlier number to the upper level position, and write the point with the later number to the lower level position. Connect the fund account points according to the adjacent account numbers in the fund destination path to obtain the hierarchical connection chart. S53: Based on the layered connection graph, call the fund round-trip segment identifier, closed segment identifier, fund destination path number, enterprise main body graph point position, natural person main body graph point position, and fund account graph point position. Extract the corresponding path number for the fund round-trip segment and the closed segment respectively, and write the fund account graph point corresponding to the path number and the fund destination connection line into the display area column to generate knowledge graph mining analysis results.

[0015] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this invention, clue fragments are constructed by co-occurring information of unit names, personnel names, and time periods in the clue text. Fund transactions are organized according to account continuity, and equity and appointments are organized according to the main body control relationship. A unified identifier set is formed by the main body name and certificate number to achieve precise alignment between clue fragments, transaction trajectories, and control relationships. Combined with the time interval filtering of synchronously appearing main body and account combinations, the clue direction and fund behavior are made consistent and mapped. On this basis, the control hierarchy transmission order and the fund flow sequence are compared to form a consistent association structure. By extracting round-trip and closed paths through the consistent first and last intervals in the account sequence, the implicit cyclic flow trajectory is presented in a structured way, enhancing the cross-level association identification ability and the clarity of path interpretation. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the acquisition process of S1 in this invention; Figure 3 This is a flowchart illustrating the acquisition process of S2 in this invention; Figure 4 This is a flowchart illustrating the acquisition process of S3 in this invention; Figure 5 This is a flowchart illustrating the acquisition process of S4 in this invention; Figure 6 This is a flowchart of the acquisition process for S5 of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the difference between them, their intended meanings are consistent. Similarly, the terms "of," "correlation (ponding)," and "correlation (ponding)" may sometimes be used interchangeably. It should be noted that, without emphasizing the difference between them, their intended meanings are consistent.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] Please see Figure 1 This invention provides a technical solution: a method for mining and analyzing knowledge graphs of clues from inspections and audits for economic and financial supervision, comprising the following steps: S1: Obtain the text of inspection and supervision clues from the inspection and supervision office terminal, as well as the fund transaction flow, corporate equity and appointment information from the data acquisition server of the regulatory data center. Segment the text of inspection and supervision clues sentence by sentence, identify the co-occurrence relationship of unit name, personnel name and time period to form clue fragments, divide the fund transaction flow by account to form continuous transaction fragments, and divide the corporate equity and appointment information by subject to form control fragments. Perform unified identification processing based on the subject name and certificate number in the clue fragments, fund transaction flow, corporate equity and appointment information, generate subject identifiers based on subject name and certificate number and form a set of subject identifiers, and establish the association relationship between subject identifiers and account identifiers and between subject identifiers and subject relationship identifiers through the account identifiers in the continuous transaction fragments and the subject relationship identifiers in the control fragments. Combine the time period in the clue fragments for filtering, determine the combination pair of account identifiers and subject identifiers that appear simultaneously in the continuous transaction fragments or control fragments within the corresponding time range, and generate the corresponding result of inspection and supervision clue funds. S2: Based on the account and entity identifiers in the fund correspondence results of inspection and supervision clues, a knowledge graph network is constructed in the graph storage device to generate enterprise entity graph points, natural person entity graph points, and fund account graph points. According to the enterprise equity and appointment information, directed connections of investment relationships are established between enterprise entity graph points and between natural person entity graph points and enterprise entity graph points, and directed connections of appointment relationships are established between natural person entity graph points and enterprise entity graph points. According to the payment account identifier and receiving account identifier in the fund transaction flow, directed connections of fund flow are established between account graph points. The descriptive content in the clue fragments formed in the inspection and supervision clue text is associated with enterprise entity graph points or natural person entity graph points and an explanation is added to form a graph relationship network with clear association relationships, and an economic and financial supervision knowledge graph network is generated. S3: Construct a network of entity control relationships based on directed connections of investment and appointment relationships in the knowledge graph network of economic and financial supervision. Starting from enterprise entity points or natural person entity points without upstream control sources, traverse enterprise entity points and natural person entity points layer by layer along the control relationship connections and form hierarchical numbers according to the control transmission path. Based on the directed connections of fund flow, connect fund account points as nodes according to fund receipt and payment relationships to form fund destination paths, and generate sequential numbers in the fund destination paths according to transaction time. Compare the hierarchical numbers of enterprise entity points and natural person entity points with the sequential numbers of fund account points to determine whether the upstream enterprise entity points or natural person entity points in the control relationship are consistent with the direction of the fund sending account points in the fund destination path. Filter enterprise entity points and natural person entity points with consistent directions and their corresponding connections, and present the hierarchical numbers and sequential numbers together to generate the control hierarchy fund sequence analysis results. S4: Based on the fund destination paths and corresponding sequence numbers obtained from the control level fund sequence analysis results, starting from the fund account map point at the starting point of the fund destination path, connect each fund account map point in sequence according to the sequence number to form an ordered account sequence. Identify fund account map points with the same account identifier in the ordered account sequence. Determine the corresponding interval segments based on the positional relationship of fund account map points with the same account identifier in the ordered account sequence. Extract the interval segments that satisfy the consistency of the first and last account identifiers. Divide the extracted interval segments into fund round-trip segments and closed segments. Correspond the occurrence of enterprise main map points or natural person main map points in the interval segments with the level number to generate the inspection and patrol cycle round-trip path results. S5: Based on the fund transfer segments, closed segments, fund destination paths, and the association between enterprise entity points, natural person entity points, and hierarchical numbers in the results of the round-trip path of the inspection and patrol, select the enterprise entity points, natural person entity points, and fund account points involved in the fund transfer segments and closed segments in the graph display device. Arrange the enterprise entity points and natural person entity points in a hierarchical manner according to the hierarchical numbers, and connect the fund destination lines between fund account points according to the fund destination paths. Display the fund destination paths corresponding to the fund transfer segments and closed segments in the graph, and generate knowledge graph mining and analysis results.

[0023] The results of the investigation and inspection clues and corresponding funds include time range identifiers, corresponding main account groups, clue association units, data matching markers, and effective combination sets. The economic and financial supervision knowledge graph network includes enterprise main node, natural person main node, fund account node, relationship connection set, and semantic description information. The results of the control level fund sequence analysis include control level sequence, fund path sequence, direction consistency marker, level correspondence relationship, and effective path set. The results of the investigation and inspection cycle path include fund round trip segments, fund closure segments, path interval identifiers, path classification markers, and path associated main entities. The results of knowledge graph mining analysis include hierarchical display structure, path display unit, key node set, relationship connection set, and analysis result view.

[0024] Please see Figure 2 The steps to obtain S1 are as follows: S11: Obtain the text of inspection and supervision clues from the inspection and supervision office terminal, as well as the fund transaction flow, corporate equity and appointment information from the data collection server of the regulatory data center. Call the period, semicolon and comma positions in the text of inspection and supervision clues to divide the sentences into segments, extract the unit name, personnel name and time period sentence by sentence, and merge them according to the number of times the same sentence appears to obtain co-occurrence segment groups. The system retrieves textual clues related to inspections and investigations from the inspection and investigation office terminal, as well as financial transaction records, corporate equity, and appointment information from the data acquisition server of the regulatory data center. It then calls an external entity data integration bus via an Ethernet interface, uses ETL (Extract, Transform, Load) tools to connect to heterogeneous databases to obtain the raw dataset, and invokes a character recognition engine to perform a full-text scan of the inspection and investigation clue text. It locates the encoded positions of all periods, semicolons, and commas in the text stream, and segments the text into independent physical segments using punctuation marks as boundaries. For each segment, it performs string matching using a pre-set entity dictionary, extracting character sequences matching the characteristics of unit names (e.g., "XXX Construction Engineering Company"), personnel names (e.g., "Li Si"), and time periods (e.g., "January to June"). The system also extracts unit names from the same segment. Personnel Name and time period Perform Cartesian product combination, count the frequency of identical entity combinations in different sentence segments, and deduplicate and merge combinations with identical company names, personnel names, and time periods. When "Li Si," "a certain construction company," and "January to June" appear simultaneously in the first and tenth sentences, aggregate these three elements into a co-occurrence tuple. Calculate the absolute offset distance of each element within a sentence segment by calling the text offset function, and calculate the number of characters between entities within the same sentence segment. (Unit: number of characters) With preset semantic association threshold Perform a numerical comparison, if If a character is selected, the combination is considered to have a significant correlation and is retained. If a character is found to be invalid discrete combination due to semantic sparsity, a removal operation is performed to preserve the core association relationship and obtain a co-occurring fragment group.

[0025] S12: Based on the co-occurrence fragment group, call the account identifier, transaction time, entity name, and document number in the fund transaction flow. Check the entity name, document number, and entity relationship identifier in the enterprise equity and appointment information item by item. Merge and mark the items with the same name and the items with the same document number. Write the account identifier and entity relationship identifier into the same entity identifier set to obtain the associated list. Based on the co-occurrence fragment group, read the account identifier from the fund transaction record. For a 19-digit account number starting with an underscore, a regular expression processor is used to extract the account number field from an XML or CSV message exported from the bank's core system. This involves reading the transaction time, entity name, and identification number (e.g., an 18-digit unified social credit code or ID card number). The system also reads the entity name, identification number, and entity relationship identifiers (e.g., legal representative, shareholder, or senior executive) from the enterprise equity and employment information. A string equality comparison is performed between the entity name in the inspection leads and the entity name in the transaction and equity information. A digit-by-digit alignment comparison is performed on the identification number. Records with the same name and identification number are identified as the same entity. All attributes of the identified entity are aggregated, and multiple bank accounts corresponding to that entity are identified. And the identifier of the main relationship in different enterprises Map them to the same logical container and assign a globally unique principal identifier to that container. If the subject Li Si corresponds to two accounts starting with 6221 and 6223, and Li Si is the financial manager of Company B, then the attribute aggregation algorithm will be used to group the relationship identifiers between these two accounts and the financial manager under the subject identifier set of Li Si. The hash mapping function will then be called to... It performs key-value pair binding with all its subordinate attributes, constructs a multi-dimensional association mapping table from all identified subject identifiers, corresponding account identifier sets, and relationship identifier sets, and obtains the association directory. S13: Based on the associated directory, call the account identifier and transaction time in the continuous transaction segment, call the subject relationship identifier and subject identifier in the control segment, perform interval comparison for the start and end time of the time period in the clue segment, select the combination of account identifier and subject identifier that falls into the same time interval and appears synchronously, and generate the corresponding result of the funds of the inspection and supervision clue. Based on the associated directory, account identifiers and the precise time of each corresponding transaction are extracted from continuous transaction segments. This time is recorded using a standard timestamp obtained through the server clock synchronization protocol (NTP). The entity identifier and its associated entity relationship identifier are retrieved from the control segment to obtain the starting point of the time period recorded in the patrol clue segment. and the end of the time period , will the transaction time and time interval Perform numerical placement determination, calculate the transaction time minus Time difference as well as Subtract the difference at the time of the transaction. The unit of measurement here is seconds (s). and If both are established, the transaction is determined to fall within the effective monitoring time interval. Simultaneously, the entity to which the account identifier belongs to the transaction occurred within that time interval is retrieved, and a set intersection operation is performed with the active entity identifiers in the control segment. Entities and accounts that have both fund transfer records and specific equity and employment relationships within a specific time period are matched. If Wang Wu used account 6225 to transfer funds to related parties between January and June, and his shareholding ratio in the target company exceeded 5% during that period, the specific mapping relationship between the entity identifier and the account identifier is recorded. The successfully matched transaction serial number is associated with the clue entry execution pointer to generate the corresponding result of the inspection and supervision clue funds.

[0026] Please see Figure 3 The steps to obtain S2 are as follows: S21: Based on the account identifier and subject identifier in the fund correspondence results of the inspection and supervision clues, call the subject identifier entries and account identifier entries in the map storage device for classification and judgment, write the subject identifier into the node sequence according to the enterprise registration mark and the natural person registration mark, write the account identifier into the account sequence according to the payment mark, and assign map point numbers to the node sequence and account sequence in sequence to establish map point catalog; Based on the account and entity identifiers in the fund matching results of inspection and supervision leads, attribute field retrieval is performed on the entity identifier entries in the storage device. The registration attribute tags corresponding to each entity identifier are extracted. Fuzzy text matching is performed between the registration attribute tags and preset character templates such as "Limited Liability Company," "Individual Business," or "Public Institution." Successfully matched entries are categorized as enterprise registration markers and written into the enterprise node sequence in the order they are read. Entries that do not match but conform to the citizen identity information verification logic are categorized as natural person registration markers and written into the natural person node sequence. Fields representing account type are extracted from the account identifiers, and those with active transaction records and high transaction frequency are identified. Accounts with more than ten transactions per quarter (in units of time / quarter) are identified based on payment and receipt characteristics and written into the account sequence. The first account to enter the enterprise node sequence... Individual subject endowment The formatted point numbering, for the first natural person node entering the natural person node sequence. Individual subject endowment The formatted point number corresponds to the first point in the account sequence. Individual accounts are granted The formatted point numbering, if the main identifier is recognized. A specific logistics company is assigned a number. The main identifier was identified. Zhao Liu was assigned a number The associated corporate account was identified. Then assign a number Establish a mapping index table between map point numbers and original physical identifiers, perform key-value pair association and storage for all generated map point numbers and original identifiers, and establish a map point catalog.

[0027] S22: Based on the map point catalog, retrieve the entries for investor name, invested entity name, employing entity name, and employing status from the enterprise equity and employment information, check the map point number and the corresponding position of the entity identifier item by item, connect the enterprise entity map points with the investment direction line, connect the natural person entity map points with the enterprise entity map points with the employment direction line, retrieve the payment account identifier and receiving account identifier from the fund transaction flow and write them into the flow direction line between the account map points to obtain the relationship network column; Based on the map data, extract the names of the investing entities from the enterprise equity information. With respect to the name of the invested entity Extract the name of the employing entity from the employment information. Company Name Locating by string hash comparison and The corresponding point number in the point catalog. and Construct a starting point between two points. The destination is The investment is directed to a specific link, and the equity ratio is used as the weight attribute of the link. If Company C invests 30% in Company D, then the following will be established: Weighted connection and ,position Corresponding natural person map point number and Corresponding enterprise drawing point number Establishing the starting point as The destination is The system links to individuals with specific job titles, such as "General Manager," and retrieves the payment account identifier for each transaction in the financial transaction history. With receiving account identifier Retrieve the corresponding account point number from the point catalog. and Establish a directed connection between two account points and obtain the total flow of funds by summing the transaction amounts between the same start and end points. ,in For the amount of a single transaction, To count the number of transactions within a given interval, the three types of connections—investment, appointment, and fund flow—are topologically linked with their corresponding node numbers to obtain a relationship network.

[0028] S23: Based on the relationship network column, call the text of the inspection and supervision clues to form the description content in the clue fragment, and attach it to the enterprise main body map point or natural person main body map point according to the main body identifier. Perform adjacency verification on the enterprise main body map point, natural person main body map point, fund account map point and the start and end numbers of each connection line, write the consecutive numbered items into the same map network column, write the description content into the corresponding description position, and generate an economic and financial supervision knowledge graph network; Based on the relationship network, retrieve clue fragments formed from the text of inspection and supervision clues. Extract the entity names and identifiers from the clue fragments, and locate the corresponding enterprise entity points through the entity identifier index map point catalog. Or natural person subject point The clue description, such as "the source and destination of funds are unknown," is attached as an unstructured text attribute to the located graph node, and the adjacency matrix of the graph database is called. Perform connectivity checks on all enterprise entity points, natural person entity points, and capital account points, and calculate the degree of each point. That is, the total number of incoming and outgoing edges connected to it, for For each graph point, its adjacent nodes within the same connected component are logically grouped and written into the same graph network column storage space. For each node's corresponding cue attachment position, offset verification is performed. The description content is then written to the corresponding node description field using pointer references. Associated with number The inspection clues are displayed in real time at the corresponding description position of the node in the graph. By calling the graphics rendering engine, the text attributes are transformed into visual labels that float above the nodes, completing the unified mapping of physical storage and logical relationship, and generating an economic and financial supervision knowledge graph network.

[0029] Please see Figure 4 The steps to obtain S3 are as follows: S31: Construct a subject control relationship network based on the directed connections of investment relationships and appointment relationships in the knowledge graph network of economic and financial supervision. Call up enterprise subject graph points, natural person subject graph points, investment relationship start and end numbers, appointment relationship start and end numbers, filter out graph points with upstream control source numbers, register the next hop graph point number layer by layer along the external connection direction of the remaining graph points, and write it into the hierarchical position according to the transmission order to establish a control layer sequence; Specifically, the following points were filtered out based on their upstream control source numbers: The corresponding positions of enterprise entity points and natural person entity points in the start and end numbers of investment relationship and employment relationship are checked. Points that appear as termination numbers in the start and end numbers of investment relationship or employment relationship and have corresponding start numbers are judged as points with upstream control source numbers and are removed from the traversal range. Registering the next-hop point number layer by layer along the external connection direction of the remaining points is as follows: Based on the connection relationship between the start and end numbers of the investment relationship and the start and end numbers of the appointment relationship, the system traverses layer by layer from the start number to the end number, writes each end number into the corresponding level position in the connection order, and merges the map point numbers that appear repeatedly in the same level position. Based on the directed connections of investment and appointment relationships in the economic and financial supervision knowledge graph network, a network of entity control relationships is constructed, and the sequence of enterprise entity graph points is invoked. Natural Person Subject Point Sequence Extract the starting point number corresponding to each investment relationship connection. With the termination point number Extract the starting point number corresponding to each employment relationship connection. With the termination point number Number all termination points and The controlled node set is compiled, and the entire graph node sequence is traversed. Each graph node number is checked to see if it exists in the controlled node set. Graph nodes that appear as termination numbers under the dimensions of investment or appointment and are simultaneously associated with corresponding start numbers are considered to have upstream control sources. Nodes with incoming edges of this type are removed from the initial traversal pool. Isolated source nodes with zero in-degree are selected as the top-level root node of the control chain. If the graph node corresponding to the natural person Sun Qi... If a node appears only as the starting number of an investment relationship and has no incoming connections, it is determined as the root node. Starting from the root node, a depth-first search topological traversal is performed along the directed edges of investment and appointment. The node number of the next hop node directly pointed to by the root node is recorded and its level is marked as level 1. The search continues to search for the downstream nodes pointed to by the first-level nodes and their level is marked as level 2. If a certain enterprise graph node... If a record is pointed to by two parent nodes, the smaller value of the level position count in different traversal paths is used to merge the records. The smaller the value, the closer it is to the core control source. Each map point is numbered according to its calculated hierarchical position. Perform key-value pair mapping to establish the control layer sequence.

[0030] S32: Based on the control layer sequence, call the payment account identifier, receiving account identifier, transaction time record and fund account plot number in the directed connection of fund flow. For the external account plot corresponding to the same payment account identifier, perform time sorting, connect the previous receiving account plot with the subsequent payment account plot in sequence, and write the order position according to the connection order. Call the enterprise main plot number, natural person main plot number and hierarchical position to compare with the fund account plot number and order position to obtain the same direction list; The specific sorting of execution time for external account points corresponding to the same payment account identifier is as follows: The transaction time records are arranged in order, and the connection between adjacent records after sorting is judged. The account points whose receiving account identifier is consistent with the paying account identifier in the subsequent records are established as continuous connection relationships and written into the order position. Based on the control layer sequence, the payment account identifier in the directed connection of fund flow is invoked. Receiving account identifier and the corresponding transaction time records Retrieve the corresponding account point number from the account point catalog. For each specific payment account identifier, all associated external transaction records are processed according to... Sort the values ​​in ascending order and extract the first value from the sorted sequence. The receiving account identifier of the record With the Payment account identifier of the record Perform a string equality comparison. If the two records match, connect the corresponding account graph points of the two transactions and calculate the time interval between the two adjacent transactions. (Unit: seconds), set time interval threshold (Corresponding to 72 hours, converted to 259,200 seconds), if This confirms a continuous funding transmission path. If the time direction consistency of the fund monitoring sequence is not established, it means that the funds have stagnated or been misappropriated at the current node. In this case, the processing logic is to interrupt the path, clear the intermediate variables, restart the new sequence counting, and assign a sequence position to each account graph point participating in the path according to the physical order of connection. Retrieve the main hierarchical position in the control layer sequence Hierarchical levels are determined by the affiliation between the entity and the account. Rank of corresponding fund accounts Perform numerical alignment and search for records where the control level increment direction and the fund flow sequence increment direction overlap under the same entity identifier to obtain a list of records with the same direction.

[0031] S33: Based on the same-direction list, call the enterprise entity map point number, natural person entity map point number, investment relationship connection number, employment relationship connection number, capital account map point number, hierarchical position, and sequential position. For items in the same column as the transmission direction and the sending direction, perform the attachment, write the hierarchical position and sequential position into the same presentation sequence, write the enterprise entity map point, natural person entity map point and corresponding connection into the associated position, and generate the control level capital sequence analysis result; Based on the list of entities in the same direction, the hierarchical position corresponding to each point is extracted by calling the enterprise entity point number, the natural person entity point number, and the associated capital contribution relationship connection number and employment relationship connection number. Sequential position corresponding to the fund account A two-dimensional coordinate mapping matrix is ​​constructed in memory, with the horizontal axis representing the direction of control transmission and the vertical axis representing the direction of fund transmission. Each item in the same-direction list is then subjected to matrix landing point verification. Through calculation Slope of the sequence and Slope of the sequence product Is it greater than zero? It was determined that the increase in control level was positively correlated with the depth of fund flow (evolution in the same direction). like If the flow is in the opposite direction or at the same level, it is determined to be a flow in the opposite direction or at the same level and is removed. Items with a product greater than zero are selected and determined to be in the same column as the transmission direction. The hierarchical positions that meet the conditions are then determined. and order of position Write the data to the same group of presentation sequences according to the same index position, extract the associated enterprise main map points, natural person main map points and the connecting objects between them, and substitute them with specific numerical instances. At the level And the order of its associated accounts is as follows: , pointing to At the level And the order of its associated accounts is as follows: If the data set is determined to conform to the logic of synchronous evolution of control and cash flow, then a strong correlation mapping is performed on the evolution relationship between the level and the sequence to generate the control level cash sequence analysis results.

[0032] Please see Figure 5 The steps to obtain S4 are as follows: S41: Based on the fund destination path and corresponding sequence number obtained from the fund sequence analysis results of the control level, call the starting fund account map point, subsequent fund account map points, and sequence number, perform the connection judgment for adjacent sequence numbers, write the fund account map points with continuous sequence and consistent connection direction into the same arrangement sequence, and record the corresponding position number of each fund account map point to establish an account arrangement list. Based on the fund flow paths and corresponding sequence numbers obtained from the control hierarchy fund sequence analysis results, extract the fund account points marked as the starting positions in the paths. And subsequent charts of various fund accounts. Retrieve the sequential numbering of each pair of adjacent map points. and Perform numerical continuity determination for adjacent numbers and calculate... and numerical difference ,like If there is a directed line connecting the two points in the chart, the connection is considered successful. The fund account points that meet the connection conditions are arranged in the same linear linked list in ascending order of their sequential numbers. Each point in the sequence is assigned a physical location number that starts from zero and increments. If the path contains a path from account X to account Y and then to account Z, and their sequential numbers are 1, 2, and 3 respectively, then in the same permutation sequence, account X is marked at position 0, account Y at position 1, and account Z at position 2. The original node attributes of each account point in the graph are retrieved, and its corresponding bank account number string is mapped and bound to the position number. The total node length of a single permutation sequence is then calculated. ,like If the sequence is not found, it will be retained for subsequent comparison. Logical grouping will be performed on all sequences that meet the conditions to create an account ranking list.

[0033] S42: Based on the account ranking list, call the account identifier and position number corresponding to each fund account point. Perform position comparison for fund account points with the same account identifier. Write the range between the previous position number and the next position number into the corresponding interval segment. Perform consistency judgment on the first account identifier and the last account identifier. Extract the consistent interval segments and register them separately according to the round-trip direction and the closed direction to obtain the segment list. Based on the account ranking list, extract the account identifier string corresponding to each capital account point in the sequence. Its position number in the sequence For all account identifiers appearing in the same sequence, perform a loop traversal search, locate the positions of any two graph points with identical account identifiers, and obtain the preceding position index. and the position number of the next position Calculate the positional spacing between two serial numbers. ,Will and All point numbers and their corresponding position numbers within the range are defined as a segment, and the first account identifier of this segment is extracted. With last account identifier Perform character-by-character alignment comparison to determine... and To determine if they point to the same bank account, perform an internal node topology search on the interval segments that are determined to be consistent, and calculate the total number of unique graph points involving different accounts within the segment. ; like If the path is from A to B and back to A, then mark this segment as a round trip and register it in the funds round trip sub-column. If the path is such that A passes through multiple different nodes such as B and C and finally returns to A, then the segment is marked as a closed path and registered in the closed sub-column. In a specific example, if the sequence is "Account M-Account N-Account O-Account M", and the account identifiers at positions 0 and 3 are consistent, and two different accounts are passed in between, then the closed segment is extracted. All the classified interval segments and their corresponding start and end numbers are stored in a structured manner to obtain the segment list.

[0034] S43: Based on the segment list, call up the enterprise main body map points, natural person main body map points, level number, and interval segment position number, perform corresponding verification for the map point appearance position in each interval segment, write the enterprise main body map points or natural person main body map points into the corresponding interval segment association position, write the level number into the same segment label position, and combine the fund round trip segment and closed segment to generate the inspection and patrol cycle round trip path result. Based on the fragment list, retrieve the enterprise entity map point sequence, the natural person entity map point sequence, and their corresponding hierarchical numbers. Retrieve the position number of each interval segment. The corresponding fund account diagram is used to locate the actual controlling entity behind the account through the ownership mapping table between the account and the entity. Number the corresponding main image points Write the associated bits of this interval segment to extract the hierarchical number assigned to the main image point in the control layer sequence. Write it to the label bit of the same segment, and perform penetration depth calculation for each segment. The calculation formula is as follows: ,in The maximum value of all subject hierarchy numbers involved in this segment. Set the penetration depth threshold to the minimum value. ,like If the segment exhibits significant cross-level complex flow characteristics, then the fund transfer segments carrying entity identifiers, level labels, and location sequence information are aggregated with the closed segments and presented synchronously according to the logical classification of transfers and closures. If it is found that account A controlled by Zhou Ba flows back to account A through account B of company C in which Zhou Ba holds shares, and Zhou Ba is at level 1 and company C is at level 2, then the calculation is... Although it did not exceed the threshold, it was still retained as a typical round-trip path. The nodes of this path and their hierarchical relationships were highlighted and associated to generate the round-trip path results of the inspection and patrol cycle.

[0035] Please see Figure 6 The steps to obtain S5 are as follows: S51: Based on the fund transfer segments, closed segments, fund destination paths, and the association between enterprise entity map points, natural person entity map points, and hierarchical numbers in the results of the round-trip path of the inspection and patrol, call the segment identifier, map point number, hierarchical number, and path number in the map display device, perform map point aggregation judgment for fund transfer segments and closed segments, and write enterprise entity map points, natural person entity map points, and fund account map points into the same display selection sequence to establish a map point selection sequence; Based on the fund transfer segments, closed segments, fund destination paths, and the correlation between enterprise entity points, natural person entity points, and hierarchical numbers in the results of the round-trip path of the inspection and supervision, extract the segment identifiers from the display equipment. Numbering of diagram points associated with each stage Hierarchical level number and path unique number For the extracted fund transfer segments and closed segments, a point aggregation judgment is performed. The first and last position indices of each interval in the segment list are traversed to obtain all active fund account points within the corresponding interval. Simultaneously retrieve the enterprise entity map points that have ownership relationships with these accounts. With natural person main point The retrieved node numbers are deduplicated and written into the same display selection sequence. The total number of nodes involved in a single segment is then calculated. Set a size determination range, if If it is determined to be a small-scale anomalous cluster, This is then identified as a complex penetration cluster. Taking a specific example, such as fragments of fund transfers... If it includes account M, account N, and their corresponding controllers, a certain person and the holding company, then... The selected sequence is stored sequentially, and the attribute fields corresponding to each segment are retrieved and marked. Attribute labels are assigned to the points contained in the round-trip path. Assign attribute labels to the points contained in the closed path. All marked point numbers and their associated metadata are encapsulated in a structured manner to create a point selection sequence.

[0036] S52: Based on the selection sequence of map points, call the map point number of the enterprise entity, the map point number of the natural person entity, the level number, and the map point number of the capital account. Sort the level number by its upper and lower positions, write the map point with the earlier number to the upper position, and write the map point with the later number to the lower position. Connect the capital account map points according to the adjacent account numbers in the capital flow path to obtain the hierarchical connection map series. Based on the selected point sequence, call the enterprise's main point number. Natural Person Main Map Point Number Hierarchical numbering Funds account point number For hierarchical numbering Perform numerical dimension sorting from smallest to largest, and set the coordinate mapping function. ,in This is the initial value for the total vertical height of the display area (in pixels). This is the hierarchical number corresponding to the entity. Calculate the vertical coordinates corresponding to each level number, based on the preset fixed level spacing reference value (in pixels). Coordinate values ​​will be used to number the levels. Smaller image points are written at the top height position of the display interface, and the layer number is assigned. Larger Write the hierarchical map points to the lower layer height position of the display interface, and set the horizontal spacing offset. The data is distributed in 200 pixel units, evenly aligned according to the total number of nodes at the same level, and the map point numbers of adjacent accounts in the fund destination path are retrieved. and Execute logical connections and extract the cumulative transaction flow from the fund transaction log. Through calculation formula Determine the width of the connecting line in pixels. Draw layered lines with directional arrows between account points and retrieve the path numbers. Aggregate the associated connected objects within the group. If it is found that the main account of level 1 transfers money to the company account of level 3, then connect the two points with a vertical height difference of 500 pixels in the interface with a directed line segment to obtain a hierarchical connected graph.

[0037] S53: Based on the hierarchical connection graph, call the fund round-trip segment identifier, closed segment identifier, fund destination path number, enterprise main body graph point position, natural person main body graph point position, and fund account graph point position. Extract the corresponding path number for fund round-trip segments and closed segments respectively, and write the fund account graph point corresponding to the path number and the fund destination connection line into the display area column to generate knowledge graph mining and analysis results. Based on the hierarchical connection diagram, the fund transfer segment identifiers stored in the attribute rendering layer are retrieved. Closed segment identifier Fund destination path number Location of main building map Location of natural person subject points Location of fund account chart For both round-trip and closed-loop segments of funds, attribute extraction based on specific path numbers is performed to filter out those related to... Associated path number and with Associated path number In the display area column, highlight masking is performed, the graphics processing unit (GPU) shader program is invoked, and the first feature rendering parameters are set. The color component control values ​​for the warning path are used to set the second feature rendering parameters. As the color component control value for the closed path, by changing the red, green, and blue (RGB) channel values ​​of the target pixel, The corresponding path is visually represented by a highly saturated warning hue, making... The corresponding paths are represented by distinctive hues (e.g., hexadecimal color values ​​#FF0000 and #800080 in this embodiment). The fund account points corresponding to the path numbers and the connecting lines indicating fund flows between them are located. The rendering state parameters are then modified by uniformly reducing the transparency of non-critical path connections to 15%, and increasing the width of critical path connections to their original calculated value. Four times that of the original, and superimposed with specific pixel raster textures to enhance the topological recognition of the path, simultaneously performing pulse diffusion animation marking on the coordinate nodes of the enterprise main graph points and the natural person main graph points, bringing in the actual demonstration loading data of the final rendering canvas, and generating results for knowledge graph mining and analysis.

[0038] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for mining and analyzing knowledge graphs of clues from inspections and audits for economic and financial supervision, characterized in that, Includes the following steps: S1: Obtain clue text, fund transaction flow and corporate equity and appointment information, segment the clue text into clue fragments, divide continuous transaction fragments and control fragments, identify the subject name and establish the association between the subject and the account and relationship identifier, combine time period to filter and match, and generate the corresponding results of the funds of the inspection and supervision clues. S2: Based on the account and entity identifiers in the fund correspondence results of the inspection and supervision clues, construct the enterprise entity, natural person entity and fund account graph points, establish the connection between investment relationship, appointment relationship and fund flow, associate the clue fragment description content, and generate an economic and financial supervision knowledge graph network; S3: Based on the connections of investment, appointment and fund flow in the economic and financial supervision knowledge graph network, construct the main control relationship network and fund destination path, compare the hierarchical number and sequence number, select the main graph points and connections with the same direction, and generate the control level fund sequence analysis results. S4: Based on the fund destination path and sequential number in the fund sequence analysis results of the control level, connect them to form an ordered account sequence, identify the same account identifier and determine the interval segment, extract the segment with the same first and last account identifier and classify the type, and generate the inspection and patrol cycle round-trip path results.

2. The method for knowledge graph mining and analysis of inspection and supervision clues for economic and financial supervision as described in claim 1, characterized in that: The results of the investigation and inspection clues and corresponding funds include time range identifiers, corresponding main account groups, clue association units, data matching markers, and effective combination sets. The economic and financial supervision knowledge graph network includes enterprise main node, natural person main node, fund account node, relationship connection set, and semantic description information. The results of the control level fund sequence analysis include control level sequence, fund path sequence, direction consistency marker, level correspondence relationship, and effective path set. The results of the investigation and inspection cyclic round-trip path include fund round-trip segments, fund closure segments, path interval identifiers, path classification markers, and path associated main entities.

3. The method for knowledge graph mining and analysis of inspection and supervision clues for economic and financial supervision as described in claim 1, characterized in that: The steps for obtaining S1 are as follows: S11: Obtain the text of inspection and supervision clues from the inspection and supervision office terminal, as well as the fund transaction flow, corporate equity and appointment information from the data collection server of the regulatory data center. Call the period, semicolon and comma positions in the text of inspection and supervision clues to divide the sentences into segments, extract the unit name, personnel name and time period sentence by sentence, and merge them according to the number of times the same sentence appears to obtain co-occurrence segment groups. S12: Based on the co-occurrence fragment group, call the account identifier, transaction time, entity name, and document number in the fund transaction flow, and check the entity name, document number, and entity relationship identifier in the enterprise equity and appointment information item by item. Merge and mark the items with the same name and the items with the same document number, and write the account identifier and entity relationship identifier into the same entity identifier set to obtain the associated list. S13: Based on the associated directory, call the account identifier and transaction time in the continuous transaction segment, call the subject relationship identifier and subject identifier in the control segment, perform interval comparison for the start and end time of the time period in the clue segment, select the combination of account identifier and subject identifier that falls in the same time interval and appears synchronously, and generate the corresponding result of the inspection and supervision clue funds.

4. The method for knowledge graph mining and analysis of inspection and supervision clues for economic and financial supervision as described in claim 1, characterized in that: The steps for obtaining S2 are as follows: S21: Based on the account identifier and subject identifier in the fund correspondence results of the inspection and investigation clues, call the subject identifier entries and account identifier entries in the map storage device for classification and judgment, write the subject identifier into the node sequence according to the enterprise registration mark and the natural person registration mark, write the account identifier into the account sequence according to the payment mark, and assign map point numbers to the node sequence and the account sequence in sequence to establish map point catalog; S22: Based on the above point catalog, retrieve the entries for investor name, invested entity name, employing entity name, and employing status from the enterprise equity and employment information, check the corresponding positions of point numbers and entity identifiers item by item, connect the enterprise entity points with the investment direction line, connect the natural person entity points with the enterprise entity points with the employment direction line, retrieve the payment account identifier and receiving account identifier from the fund transaction flow and write them into the flow direction line between the account points to obtain the relationship network column; S23: Based on the aforementioned relationship network column, call the inspection and supervision clue text to form the description content in the clue fragment, and attach it to the enterprise main body map point or natural person main body map point according to the main body identifier. Perform adjacency verification on the enterprise main body map point, natural person main body map point, fund account map point and the start and end numbers of each connecting line, write the consecutive numbered items into the same map network column, write the description content into the corresponding description position, and generate an economic and financial supervision knowledge graph network.

5. The method for knowledge graph mining and analysis of inspection and supervision clues for economic and financial supervision as described in claim 1, characterized in that: The steps for obtaining S3 are as follows: S31: Construct a subject control relationship network based on the directed connections of investment relationships and appointment relationships in the economic and financial supervision knowledge graph network. Call up enterprise subject graph points, natural person subject graph points, investment relationship start and end numbers, appointment relationship start and end numbers, filter out graph points with upstream control source numbers, register the next hop graph point number layer by layer along the external connection direction of the remaining graph points, and write it into the hierarchical position according to the transmission order to establish a control layer sequence; S32: Based on the control layer sequence, call the payment account identifier, receiving account identifier, transaction time record and fund account point number in the directed connection of fund flow, perform time sorting for the external account points corresponding to the same payment account identifier, connect the previous receiving account point and the subsequent payment account point in sequence, and write the order position according to the connection order, call the enterprise main point number, natural person main point number and hierarchical position to compare with the fund account point number and order position to obtain the same direction list; S33: Based on the aforementioned list of parallel paths, call the enterprise entity map point number, natural person entity map point number, investment relationship connection number, employment relationship connection number, capital account map point number, hierarchical position, and sequential position. For items in the same column as the transmission direction and the sending direction, perform the attachment, write the hierarchical position and sequential position into the same presentation sequence, write the enterprise entity map point, natural person entity map point and corresponding connection into the associated position, and generate the control level capital sequence analysis result.

6. The method for knowledge graph mining and analysis of inspection and supervision clues for economic and financial supervision as described in claim 5, characterized in that: The specific steps for filtering out points with upstream control source numbers are as follows: The corresponding positions of the enterprise entity map points and the natural person entity map points in the start and end numbers of the investment relationship and the start and end numbers of the employment relationship are checked. Map points that appear as termination numbers in the start and end numbers of the investment relationship or the start and end numbers of the employment relationship and have corresponding start numbers are determined to be map points with upstream control source numbers and are removed from the traversal range. The specific steps for registering the next-hop point number layer by layer along the external connection direction of the remaining point are as follows: Based on the connection relationship between the start and end numbers of the investment relationship and the start and end numbers of the appointment relationship, a layer-by-layer traversal is performed, and each termination number is written into the corresponding level position in the connection order. The map point numbers that appear repeatedly in the same level position are merged and recorded.

7. The method for knowledge graph mining and analysis of inspection and supervision clues for economic and financial supervision as described in claim 5, characterized in that: The specific sorting of execution time for external account points corresponding to the same payment account identifier is as follows: The transaction time records are arranged sequentially, and the connection between adjacent records after sorting is judged. Account points whose receiving account identifiers are consistent with the paying account identifiers in subsequent records are established as continuous connections and written into the order position.

8. The method for knowledge graph mining and analysis of inspection and supervision clues for economic and financial supervision as described in claim 1, characterized in that: The steps for obtaining S4 are as follows: S41: Based on the fund destination path and corresponding sequence number obtained from the fund sequence analysis results of the control level, call the starting fund account map point, subsequent fund account map points, and sequence number, perform a connection judgment for adjacent sequence numbers, write the fund account map points with continuous sequence and consistent connection direction into the same arrangement sequence, and record the corresponding position number of each fund account map point to establish an account arrangement list. S42: Based on the account ranking list, call the account identifier and position number corresponding to each fund account point, perform position comparison for fund account points with the same account identifier, write the range between the previous position number and the next position number into the corresponding interval segment, perform consistency judgment on the first account identifier and the last account identifier, extract the consistent interval segments and register them separately according to the round-trip direction and the closed direction to obtain the segment list; S43: Based on the fragment list, call up the enterprise main body map points, natural person main body map points, level number, and interval fragment position number, perform corresponding verification for the map point appearance position in each interval fragment, write the enterprise main body map points or natural person main body map points into the corresponding interval fragment association position, write the level number into the same fragment label position, and combine the fund round trip fragments and closed fragments to generate the inspection and patrol cycle round trip path result.

9. The method for knowledge graph mining and analysis of inspection and supervision clues for economic and financial supervision as described in claim 1, characterized in that: The method further includes: S5: Based on the fund round-trip segments, closed segments, fund destination paths and main map point hierarchical numbers in the inspection and patrol cycle round-trip path results, select relevant map points and arrange them in layers, connect the fund destination lines, display the corresponding fund destination paths respectively, and generate knowledge graph mining analysis results; The knowledge graph mining and analysis results include a hierarchical display structure, path display units, key node sets, relationship connection sets, and analysis result views.

10. The method for knowledge graph mining and analysis of inspection and supervision clues for economic and financial supervision as described in claim 9, characterized in that: The steps for obtaining S5 are as follows: S51: Based on the fund transfer segments, closed segments, fund destination paths, and the association between enterprise entity map points, natural person entity map points, and hierarchical numbers in the results of the patrol and inspection cycle, call the segment identifier, map point number, hierarchical number, and path number in the map display device, perform map point aggregation judgment for fund transfer segments and closed segments, and write enterprise entity map points, natural person entity map points, and fund account map points into the same display selection sequence to establish a map point selection sequence; S52: Based on the selected point sequence, call the enterprise main point number, natural person main point number, level number, and fund account point number. Sort the level number by position, write the point with the earlier number to the upper level position, and write the point with the later number to the lower level position. Connect the fund account points according to the adjacent account numbers in the fund destination path to obtain the hierarchical connection chart. S53: Based on the layered connection graph, call the fund round-trip segment identifier, closed segment identifier, fund destination path number, enterprise main body graph point position, natural person main body graph point position, and fund account graph point position. Extract the corresponding path number for the fund round-trip segment and the closed segment respectively, and write the fund account graph point corresponding to the path number and the fund destination connection line into the display area column to generate knowledge graph mining analysis results.

Images