A method and device for identifying tax-related risk enterprises
The method and device enhance the efficiency and accuracy of identifying tax-related risk enterprises by using a directed enterprise association graph and label propagation to analyze transaction data, effectively identifying risk enterprises with reduced manual effort.
Patent Information
- Application Number
- CN202011060394.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-09-30
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2040-09-30
AI Technical Summary
The existing methods for identifying tax-related risk enterprises are inefficient and inaccurate, consuming significant time and manpower while relying on uncertain empirical judgments.
A method and device that utilize a directed enterprise association graph to identify tax-related risk enterprises by determining the maximum connected subgraph of a specified abnormal enterprise, calculating edge weights based on transaction data, and predicting labels using label propagation algorithms to identify unlabeled enterprises as risk or credit enterprises.
This approach significantly improves the efficiency and accuracy of identifying tax-related risk enterprises by quickly identifying associated risks through a graph-based analysis with low computational complexity.
Smart Images

Figure CN114358893B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of information security technology, and in particular to a method and device for identifying tax-related risky enterprises. Background Art
[0002] With the full launch of the Golden Tax Phase III tax management information system, the era of tax big data has arrived. The Golden Tax Phase III tax management information system collects a large amount of taxpayer identity information, asset information, invoicing information and other information, making taxpayers' business operations more transparent.
[0003] At present, the inspection departments of the national and local tax bureaus mainly monitor tax-related abnormal enterprises through project analysis, indicator analysis, and manual experience. This not only consumes a lot of time and manpower costs, but also the basis for empirical judgment is uncertain. Therefore, the efficiency and accuracy of identifying whether enterprises associated with tax-related abnormal enterprises also have tax-related risks are low.
[0004] Therefore, how to improve the efficiency and accuracy of identifying tax-related risk enterprises based on known tax-related abnormal enterprises is one of the technical problems that need to be urgently solved in the existing technology. Summary of the invention
[0005] In order to solve the problem of low efficiency and accuracy in identifying tax-related risk enterprises associated with abnormal tax-related enterprises in the prior art, an embodiment of the present invention provides a method and device for identifying tax-related risk enterprises.
[0006] In a first aspect, an embodiment of the present invention provides a method for identifying tax-related risk enterprises, comprising:
[0007] In response to a trigger request for querying tax-risk enterprises, a maximum connected subgraph of a specified tax-risk abnormal enterprise is obtained from a stored enterprise association graph, wherein the enterprise association graph is a directed association graph constructed based on the transaction relationship between enterprises, with the first transaction party as a starting node, the second transaction party as an end node, and the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party as directed edges, and the maximum connected subgraph represents the maximum set having a connected relationship with the specified tax-risk abnormal enterprise;
[0008] For each pair of the first transaction party and the second transaction party in the maximum connected subgraph, determine the weight of the directed edge from the node corresponding to the first transaction party in the pair of the first transaction party and the second transaction party to the node corresponding to the second transaction party according to the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party in the pair of the first transaction party and the second transaction party, wherein the weight of the directed edge represents the closeness of the relationship between the first transaction party and the second transaction party;
[0009] Determine the propagation probability between every two nodes in the maximum connected subgraph according to the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party;
[0010] According to the label marked for the designated tax-related abnormal enterprise, the labels marked for several other designated enterprises, and the propagation probability between every two nodes, predict the label of the unlabeled enterprise, wherein the label is the probability that the enterprise is a tax-related abnormal enterprise;
[0011] According to the predicted label of the unlabeled enterprise, it is identified whether the unlabeled enterprise is a tax risk enterprise.
[0012] In the tax risk enterprise identification method provided in the embodiment of the present invention, a computer device responds to a trigger request for querying tax risk enterprises, obtains a maximum connected subgraph of a specified tax-related abnormal enterprise from a stored enterprise association graph, the enterprise association graph is a directed association graph constructed based on the transaction relationship between enterprises, with the first transaction party as the starting node, the second transaction party as the ending node, and the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party as directed edges, the maximum connected subgraph represents the maximum set that has a connected relationship with the specified tax-related abnormal enterprise, for each pair of the first transaction party and the second transaction party in the maximum connected subgraph, according to the total amount of transaction invoices from the first transaction party to the second transaction party in the pair of the first transaction party and the second transaction party The amount and number of transactions are used to determine the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in the pair of the first transaction party and the second transaction party, wherein the weight of the directed edge represents the closeness of the relationship between the first transaction party and the second transaction party. Furthermore, according to the determined weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party, the propagation probability between every two nodes in the maximum connected subgraph is determined. According to the labels marked for the designated tax-related abnormal enterprises, the labels marked for several other designated enterprises, and the propagation probability between every two nodes, the labels of the enterprises without labels are predicted, wherein the label is the probability that the enterprise is a tax-related abnormal enterprise. Furthermore, according to the predicted The labels of unlabeled enterprises are measured to identify whether the unlabeled enterprises are tax-risk enterprises. Compared with the prior art, in the embodiment of the present invention, a directed enterprise association graph is constructed according to the transaction relationship between enterprises to record the association relationship between enterprises. By finding the largest connected subgraph of known tax-related abnormal enterprises in the enterprise association graph, other enterprises that have an association relationship with the known tax-related abnormal enterprises can be quickly found. Then, according to the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party in each pair of the first transaction party and the second transaction party in the largest connected subgraph, the weight of the directed edge between the starting node corresponding to the first transaction party in each pair of the first transaction party and the second transaction party and the ending node corresponding to the second transaction party is determined as a representation The method measures the closeness of the relationship between the first transaction party and the second transaction party, and determines the propagation probability between each two nodes in the maximum connected subgraph according to the weight of the edge between the starting node corresponding to the first transaction party and the ending node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party. Based on the propagation probability between each two nodes and the labels of the labeled enterprises, the labels of the unlabeled enterprises are predicted according to the label propagation algorithm, that is, the probability of predicting that the unlabeled enterprises are tax-related abnormal enterprises is predicted, so as to identify whether the unlabeled enterprises are tax-related risk enterprises. The algorithm has low complexity and good classification effect, which saves a lot of time for subsequent judgment based on manual experience and analysis on whether enterprises associated with known tax-related abnormal enterprises have tax-related risks.Improved the efficiency and accuracy of identifying whether enterprises related to known tax-related abnormal enterprises are tax-related risk enterprises.
[0013] Preferably, for each pair of the first transaction party and the second transaction party in the maximum connected subgraph, according to the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party in the pair of the first transaction party and the second transaction party, the weight of the directed edge from the node corresponding to the first transaction party in the pair of the first transaction party and the second transaction party to the node corresponding to the second transaction party is determined, specifically including:
[0014] For each pair of the first transaction party and the second transaction party in the largest connected subgraph, pre-processing the total amount of transaction invoices from the first transaction party to the second transaction party and the number of transactions respectively;
[0015] Normalizing the pre-processed total amount of the transaction invoices from the first transaction party to the second transaction party and the pre-processed number of transactions respectively;
[0016] The weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party is determined according to the normalized preprocessed total amount of transaction invoices from the first transaction party to the second transaction party and the normalized preprocessed number of transactions.
[0017] Preferably, pre-processing the total amount of the transaction invoice from the first transaction party to the second transaction party specifically includes:
[0018] The total amount of the transaction invoice from the first transaction party to the second transaction party is pre-processed using the following formula:
[0019]
[0020] in, Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party after pre-processing;
[0021] f se Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party;
[0022] fpje means fpje se max(fpje) represents the maximum value of the total amount of transaction invoices generated by all transaction parties in the maximum connected subgraph; and
[0023] The number of transactions from the first transaction party to the second transaction party is preprocessed using the following formula:
[0024]
[0025] in, Indicates the number of transactions from the first transaction party to the second transaction party after preprocessing;
[0026] count se Indicates the number of transactions from the first transaction party to the second transaction party;
[0027] count means count se max(count) represents the maximum value of the total number of transactions generated by all transaction parties in the maximum connected subgraph.
[0028] Preferably, the pre-processed total amount of the transaction invoice from the first transaction party to the second transaction party is normalized, specifically including:
[0029] The total amount of the pre-processed transaction invoice from the first transaction party to the second transaction party is normalized by the following formula:
[0030]
[0031] in, represents the normalized pre-processed total amount of the transaction invoice from the first transaction party to the second transaction party; and
[0032] The number of transactions from the first transaction party to the second transaction party after preprocessing is normalized by the following formula:
[0033]
[0034] in, Indicates the normalized preprocessed number of transactions from the first transaction party to the second transaction party.
[0035] Preferably, determining the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party according to the normalized preprocessed total amount of the transaction invoices from the first transaction party to the second transaction party and the normalized preprocessed number of transactions, specifically includes:
[0036] The weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party is determined by the following formula:
[0037]
[0038] Wherein, ω represents the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party.
[0039] Preferably, determining the propagation probability between every two nodes in the maximum connected subgraph according to the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party, specifically includes:
[0040] The propagation probability between every two nodes in the maximum connected subgraph is determined by the following formula:
[0041]
[0042] Among them, T ij represents the transmission probability from node j to node i;
[0043] ω ij Represents the weight of the directed edge from node i to node j;
[0044] n is the total number of nodes in the largest connected subgraph.
[0045] Preferably, the label of the unlabeled enterprise is predicted based on the label labeled for the designated tax-related abnormal enterprise, the label labeled for other designated enterprises, and the propagation probability between every two nodes, specifically including:
[0046] Define the matrix Y∈(l+u)*C={Y i,C}, where l is the number of labeled enterprises, u is the number of unlabeled enterprises, C represents the category matrix to which the enterprise belongs, C = {1,0}, when C = 1, it represents an abnormal tax-related enterprise, when C = 0, it represents a tax-related credit enterprise, Y i,C represents the probability that node i belongs to category C;
[0047] A propagation probability matrix is formed according to the propagation probability between each two nodes; and label propagation is performed according to the following steps until convergence:
[0048] Each node adds the label values propagated by the surrounding nodes according to the propagation probability from its surrounding nodes to the node, and updates it to its own probability distribution to obtain Y t =T×Y t-1 , where t≥1, Y t represents the matrix Y after the tth update, Y t-1 Indicates the matrix Y before the t-th update;
[0049] Reset the labels of the labeled enterprises in the matrix Y, limit the labeled data, and reassign the probability distribution of the labeled data to the initial value.
[0050] Preferably, according to the predicted label of the unlabeled enterprise, identifying whether the unlabeled enterprise is a tax risk enterprise specifically includes:
[0051] When it is determined that the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is greater than or equal to a preset threshold, the unlabeled enterprise is determined to be a tax-related risk enterprise;
[0052] When it is determined that the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is less than the preset threshold, the unlabeled enterprise is determined to be a tax-related credit enterprise.
[0053] In a second aspect, an embodiment of the present invention provides a device for identifying tax-related risky enterprises, comprising:
[0054] an acquisition unit, configured to respond to a trigger request for querying tax-risk enterprises and acquire a maximum connected subgraph of a specified tax-related abnormal enterprise from a stored enterprise association graph, wherein the enterprise association graph is a directed association graph constructed based on a transaction relationship between enterprises, with the first transaction party as a starting node, the second transaction party as an end node, and the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party as directed edges, wherein the maximum connected subgraph represents the maximum set having a connected relationship with the specified tax-related abnormal enterprise;
[0055] A first determining unit is configured to determine, for each pair of first transaction parties and second transaction parties in the maximum connected subgraph, a weight of a directed edge from a node corresponding to the first transaction party in the pair of first transaction parties to a node corresponding to the second transaction party in the pair of first transaction parties and second transaction parties according to a total amount of transaction invoices and a number of transactions from the first transaction party to the second transaction party in the pair of first transaction parties, wherein the weight of the directed edge represents a closeness of a relationship between the first transaction party and the second transaction party;
[0056] A second determining unit, configured to determine a propagation probability between each two nodes in the maximum connected subgraph according to the weight of a directed edge from a node corresponding to the first transaction party to a node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party;
[0057] A label prediction unit, used to predict the label of an unlabeled enterprise based on the label labeled for the designated tax-related abnormal enterprise, the labels labeled for several other designated enterprises, and the propagation probability between every two nodes, wherein the label is the probability that the enterprise is a tax-related abnormal enterprise;
[0058] The identification unit is used to identify whether the unlabeled enterprise is a tax-related risk enterprise based on the predicted label of the unlabeled enterprise.
[0059] Preferably, the first determination unit is specifically used to pre-process the total amount of transaction invoices from the first transaction party to the second transaction party and the number of transactions for each pair of the first transaction party and the second transaction party in the maximum connected subgraph; normalize the pre-processed total amount of transaction invoices from the first transaction party to the second transaction party and the number of transactions after pre-processing; and determine the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party based on the normalized total amount of transaction invoices from the first transaction party to the second transaction party and the normalized number of transactions after pre-processing.
[0060] Preferably, the first determining unit is specifically configured to pre-process the total amount of the transaction invoice from the first transaction party to the second transaction party by using the following formula:
[0061]
[0062] in, Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party after pre-processing;
[0063] f se Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party;
[0064] fpje means fpje se max(fpje) represents the maximum value of the total amount of transaction invoices generated by all transaction parties in the maximum connected subgraph; and
[0065] The number of transactions from the first transaction party to the second transaction party is preprocessed using the following formula:
[0066]
[0067] in, Indicates the number of transactions from the first transaction party to the second transaction party after preprocessing;
[0068] count se Indicates the number of transactions from the first transaction party to the second transaction party;
[0069] count means count se max(count) represents the maximum value of the total number of transactions generated by all transaction parties in the maximum connected subgraph.
[0070] Preferably, the first determining unit is specifically used to normalize the pre-processed total amount of the transaction invoice from the first transaction party to the second transaction party by using the following formula:
[0071]
[0072] in, represents the normalized pre-processed total amount of the transaction invoice from the first transaction party to the second transaction party; and
[0073] The number of transactions from the first transaction party to the second transaction party after preprocessing is normalized by the following formula:
[0074]
[0075] in, Indicates the normalized preprocessed number of transactions from the first transaction party to the second transaction party.
[0076] Preferably, the first determining unit is specifically configured to determine the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party by the following formula:
[0077]
[0078] Wherein, ω represents the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party.
[0079] Preferably, the second determining unit is specifically configured to determine the propagation probability between every two nodes in the maximum connected subgraph by using the following formula:
[0080]
[0081] Among them, T ij represents the transmission probability from node j to node i;
[0082] ω ij Represents the weight of the directed edge from node i to node j;
[0083] n is the total number of nodes in the largest connected subgraph.
[0084] Preferably, the label prediction unit is specifically used to define a matrix Y∈(l+u)*C={Y i,C}, where l is the number of labeled enterprises, u is the number of unlabeled enterprises, C represents the category matrix to which the enterprise belongs, C = {1,0}, when C = 1, it represents an abnormal tax-related enterprise, when C = 0, it represents a tax-related credit enterprise, Y i,C represents the probability that node i belongs to category C;
[0085] A propagation probability matrix is formed according to the propagation probability between each two nodes; and label propagation is performed according to the following steps until convergence:
[0086] Each node adds the label values propagated by the surrounding nodes according to the propagation probability from its surrounding nodes to the node, and updates it to its own probability distribution to obtain Y t =T×Y t-1 , where t≥1, Y t represents the matrix Y after the tth update, Y t-1 Indicates the matrix Y before the t-th update;
[0087] Reset the labels of the labeled enterprises in the matrix Y, limit the labeled data, and reassign the probability distribution of the labeled data to the initial value.
[0088] Preferably, the identification unit is specifically used to determine that the unlabeled enterprise is a tax-related risk enterprise when the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is greater than or equal to a preset threshold; and to determine that the unlabeled enterprise is a tax-related credit enterprise when the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is less than the preset threshold.
[0089] The technical effects of the tax risk enterprise identification device provided by the present invention can be referred to the technical effects of the above-mentioned first aspect or each implementation method of the first aspect, and will not be repeated here.
[0090] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the method for identifying tax-risk enterprises of the present invention is implemented when the processor executes the program.
[0091] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for identifying tax-risk enterprises described in the present invention.
[0092] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0093] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:
[0094] Figure 1 A schematic diagram of an application scenario of the tax risk enterprise identification method provided by an embodiment of the present invention;
[0095] Figure 2 A schematic diagram of the implementation process of the tax risk enterprise identification method provided by the embodiment of the present invention;
[0096] Figure 3 A schematic diagram of the implementation process of constructing an enterprise association graph provided by an embodiment of the present invention;
[0097] Figure 4 An exemplary enterprise association diagram provided for an embodiment of the present invention;
[0098] Figure 5 A schematic diagram of an implementation flow of determining the weight of a directed edge from a node corresponding to the seller to a node corresponding to the buyer in each pair of sellers and buyers in a maximum connected subgraph provided by an embodiment of the present invention;
[0099] Figure 6 It is a schematic diagram of the Sigmoid function image;
[0100] Figure 7 A schematic diagram of an implementation process for identifying whether an unlabeled enterprise is a tax risk enterprise provided by an embodiment of the present invention;
[0101] Figure 8 A schematic diagram of the structure of a device for identifying tax-related risk enterprises provided by an embodiment of the present invention;
[0102] Fig. 9 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0103] In order to solve the problem of low efficiency and accuracy in identifying tax-related risk enterprises associated with abnormal tax-related enterprises in the prior art, an embodiment of the present invention provides a method and device for identifying tax-related risk enterprises.
[0104] First reference Figure 1 , which is a schematic diagram of an application scenario of the method for identifying tax-related risk enterprises provided in an embodiment of the present invention, and may include a terminal 110 and a computer device 120. When it is necessary to query a tax-related risk enterprise, for example, when the computer device 120 receives a query request for a tax-related risk enterprise sent by the terminal 110, the computer device 120 may return the identified tax-related risk enterprise to the terminal 110 according to the query request for the tax-related risk enterprise. In another application scenario, the computer device 120 may also automatically trigger a query of a tax-related risk enterprise. For example, the computer device 120 may be set to automatically execute the steps of the method for identifying tax-related risk enterprises provided in an embodiment of the present invention at preset intervals to identify tax-related risk enterprises; or a request for querying a tax-related risk enterprise may be triggered by a manager, and the computer device 120 responds to the trigger request and executes the steps of the method for identifying tax-related risk enterprises provided in an embodiment of the present invention to identify tax-related risk enterprises.
[0105] The computer device 120 may be an independent physical server or terminal, or a cloud server that provides basic cloud computing services such as cloud servers, cloud databases, and cloud storage. The terminal 110 may be, but is not limited to, a smart phone, a tablet computer, a laptop computer, a desktop computer, etc. The computer device 120 and the terminal 110 may be connected via a network, which is not limited in the embodiment of the present invention.
[0106] Based on the above application scenarios, the following will refer to the attached Figure 2 to Figure 6 The exemplary embodiments of the present invention are described in more detail. It should be noted that the above application scenarios are only shown to facilitate understanding of the spirit and principle of the present invention, and the embodiments of the present invention are not limited in any way. On the contrary, the embodiments of the present invention can be applied to any applicable scenario.
[0107] The preferred embodiments of the present invention are described below in conjunction with the drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention. In addition, the embodiments of the present invention and the features in the embodiments may be combined with each other if there is no conflict.
[0108] like Figure 2 As shown, it is a schematic diagram of the implementation process of the tax risk enterprise identification method provided by an embodiment of the present invention. The tax risk enterprise identification method can be applied to the above-mentioned computer device 120 and may include the following steps:
[0109] S11. In response to a trigger request for querying tax-related risk enterprises, the maximum connected subgraph of the specified tax-related abnormal enterprise is obtained from the stored enterprise association graph, wherein the enterprise association graph is a directed association graph constructed based on the transaction relationship between enterprises, with the first transaction party as the starting node, the second transaction party as the ending node, and the total transaction invoice amount and the number of transactions from the first transaction party to the second transaction party as directed edges.
[0110] In a specific implementation, the computer device responds to a trigger request for querying tax-risk enterprises, and obtains the maximum connected subgraph of the specified tax-related abnormal enterprise from the stored enterprise association graph, wherein the maximum connected subgraph represents the maximum set that has a connected relationship with the specified tax-related abnormal enterprise, the first transaction party is the seller, and the second transaction party is the buyer.
[0111] In the embodiment of the present invention, an abnormal tax-related enterprise may be an enterprise that issues abnormal invoices, or an enterprise that uses abnormal invoices to offset taxes. Invoices that fail to pass tax returns or have completed tax returns but are being investigated by the tax authorities can all be considered abnormal invoices, and the embodiment of the present invention does not limit this. A tax-related risk enterprise refers to an enterprise that has a certain probability of issuing abnormal invoices or using abnormal invoices to offset taxes.
[0112] Specifically, the computer device is as follows: Figure 3 The process shown in the figure builds an enterprise association diagram, which can include the following steps:
[0113] S21. Acquire enterprise transaction data, where the enterprise transaction data includes the total amount of transaction invoices from the seller to the buyer and the number of transactions.
[0114] In specific implementation, the computer device obtains enterprise transaction data from the Golden Tax Phase III tax management information system, and the enterprise transaction data may include but is not limited to the following data: enterprise information, the total amount of transaction invoices from the seller to the buyer, and the number of transactions from the seller to the buyer (i.e., the total number of transactions from the seller to the buyer), wherein the enterprise information includes information such as the enterprise name and enterprise code. The invoice in the embodiment of the present invention may be but is not limited to a value-added tax invoice.
[0115] S22. According to the transaction relationship between enterprises, an enterprise association graph is constructed with the seller as the starting node, the buyer as the ending node, and the total transaction invoice amount and transaction number from the seller to the buyer as directed edges.
[0116] During specific implementation, the computer equipment constructs a directed enterprise association graph based on the purchase and sales relationship between enterprises, with the seller as the starting node, the buyer as the ending node, and the total transaction invoice amount and transaction number from the seller to the buyer as directed edges.
[0117] Specifically, you can use start_point to indicate the starting node; use end_point to indicate the ending node; use fpje se Indicates the total amount of transaction invoices from the starting node (start_point) to the ending node (end_point), that is, the total amount of transaction invoices from the seller to the buyer, where s is the abbreviation of the seller (start_point) and e is the abbreviation of the buyer (end_point); use count se Indicates the number of transactions from the start node to the end node, that is, the total number of transactions from the seller to the buyer.
[0118] like Figure 4 As shown, it is an exemplary enterprise association diagram provided by an embodiment of the present invention. The total amount of transaction invoices from the seller to the buyer and the number of transactions between every two enterprises with a purchase and sale relationship are shown in Table 1:
[0119] Table 1
[0120] start_point end_point f count Company A Company B 2880 67 Company A C Enterprise 1550 39 Company B G Enterprise 3310 50 Company B H Enterprise 4790 88 C Enterprise Company A 4550 1 C Enterprise E-Enterprise 2990 21 C Enterprise F Enterprise 2800 10 D Enterprise Company A 5080 95 D Enterprise K Enterprise 8850 98 E-Enterprise F Enterprise 2430 72 E-Enterprise D Enterprise 4030 73 G Enterprise H Enterprise 9090 76 G Enterprise F Enterprise 7330 100 H Enterprise I Enterprise 8990 42 H Enterprise J Enterprise 8950 22 K Enterprise J Enterprise 7920 59 M Enterprise N Enterprise 1000 100 O Enterprise M Enterprise 2000 20 O Enterprise N Enterprise 3000 37
[0121] Depend on Figure 4It can be seen that the enterprise association graph is a directed association graph consisting of the seller pointing to the buyer, as shown in Table 1. Taking Enterprise A and Enterprise C as an example, when Enterprise A is the seller and Enterprise C is the buyer, the total amount of transaction invoices from Enterprise A to Enterprise C is 1,550 yuan, and the number of transactions from Enterprise A to Enterprise C is 39 times. When Enterprise C is the seller and Enterprise A is the buyer, the total amount of transaction invoices from Enterprise C to Enterprise A is 4,550 yuan, and the number of transactions from Enterprise C to Enterprise A is 1 time.
[0122] Assumptions Figure 4 Enterprise A in the figure is the designated tax-related abnormal enterprise, that is, Enterprise A is a known tax-related abnormal enterprise. The largest connected subgraph of Enterprise A is the connection graph of other enterprises except Enterprise M, Enterprise N and Enterprise O, that is, the connected subgraph composed of Enterprise A, Enterprise B, Enterprise C, Enterprise D, Enterprise E, Enterprise F, Enterprise G, Enterprise H, Enterprise I, Enterprise J and Enterprise K. In the largest connected subgraph of Enterprise A, all enterprises except Enterprise A have direct or indirect connections with Enterprise A.
[0123] S12. For each pair of the first transaction party and the second transaction party in the maximum connected subgraph, determine the weight of the directed edge from the node corresponding to the first transaction party in the pair of the first transaction party and the second transaction party to the node corresponding to the second transaction party according to the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party in the pair of the first transaction party and the second transaction party.
[0124] During specific implementation, the computer device determines the weight of the directed edge from the (starting) node corresponding to the seller to the (ending) node corresponding to the buyer in each pair of sellers and buyers in the largest connected subgraph according to the total amount of transaction invoices and the number of transactions from the seller to the buyer in the pair of sellers and buyers, where the weight of the edge represents the closeness of the relationship between the seller and the buyer.
[0125] Specifically, according to Figure 5 The steps shown determine the weight of the directed edge from the node corresponding to the seller to the node corresponding to the buyer in each pair of sellers and buyers in the maximum connected subgraph, which may include the following steps:
[0126] S31. For each pair of sellers and buyers in the largest connected subgraph, pre-process the total amount of transaction invoices from the seller to the buyer and the number of transactions.
[0127] In specific implementation, for each pair of sellers and buyers in the largest connected subgraph, the total amount of transaction invoices from the seller to the buyer is preprocessed using the following formula:
[0128]
[0129] in, It represents the total amount of the transaction invoice between the seller and the buyer after pre-processing, s represents the abbreviation of the seller (i.e. the first transaction party) (start_point), and e represents the abbreviation of the buyer (i.e. the second transaction party) (end_point);
[0130] f se Indicates the total amount of the transaction invoice from the seller to the buyer;
[0131] fpje means fpje se max(fpje) represents the maximum value in the total amount of transaction invoices generated by all transaction parties in the maximum connected subgraph, that is, max(fpje) represents the maximum value in the total amount of transaction invoices from the seller to the buyer in each pair of sellers and buyers in the maximum connected subgraph.
[0132] The transaction times from the seller to the buyer are preprocessed by the following formula:
[0133]
[0134] in, Indicates the number of transactions from the first transaction party to the second transaction party after preprocessing;
[0135] count se Indicates the number of transactions from the first transaction party to the second transaction party;
[0136] count means count se max(count) represents the maximum value of the total number of transactions generated by all transaction parties in the maximum connected subgraph, that is, max(count) represents the maximum value of the number of transactions from the seller to the buyer in each pair of sellers and buyers in the maximum connected subgraph.
[0137] In specific implementation, due to different sellers and buyers with transaction relationships, the data of the number of transactions from the seller to the buyer and the total amount of transaction invoices are quite different. In the embodiment of the present invention, the Sigmoid function is used to normalize the number of transactions and the total amount of transaction invoices respectively. The Sigmoid function has the characteristics of strict monotonicity, central symmetry with (0, 0.05), and a value range of (0, 1). The Sigmoid function is specifically as follows:
[0138]
[0139] The image of Sigmoid function is as follows Figure 6As shown, when the x value exceeds the range of [-5, 5], the function value f(x) basically does not change, and the values are very close. Therefore, in the present invention, the range of x is controlled within [-5, 5]. Therefore, before the number of transactions from the seller to the buyer and the total amount of transaction invoices from the seller to the buyer are normalized respectively, the number of transactions from the seller to the buyer and the total amount of transaction invoices from the seller to the buyer are preprocessed as described above, so that the value ranges of the number of transactions from the seller to the buyer and the total amount of transaction invoices from the seller to the buyer are controlled within [-5, 5]. Then, the total amount of transaction invoices from the seller to the buyer and the number of transactions from the seller to the buyer after preprocessing are normalized according to the Sigmoid function.
[0140] S32, respectively normalizing the pre-processed total amount of the transaction invoice from the seller to the buyer and the pre-processed number of transactions.
[0141] In specific implementation, the total amount of the transaction invoice from the seller to the buyer after pre-processing is normalized by the following formula:
[0142]
[0143] in, Represents the normalized pre-processed total transaction invoice amount from the seller to the buyer.
[0144] The preprocessed transaction times from seller to buyer are normalized by the following formula:
[0145]
[0146] in, Represents the normalized number of transactions from seller to buyer after preprocessing.
[0147] S33. Determine the weight of the directed edge from the node corresponding to the seller to the node corresponding to the buyer based on the normalized pre-processed total amount of the transaction invoices from the seller to the buyer and the normalized pre-processed number of transactions.
[0148] In specific implementation, for each pair of sellers and buyers in the maximum connected subgraph, the weight of the directed edge from the node corresponding to the seller to the node corresponding to the buyer in the seller-buyer pair is determined by the following formula:
[0149]
[0150] Wherein, ω represents the weight of the directed edge from the node corresponding to the seller to the node corresponding to the buyer.
[0151] Continuing with the above example, as shown in Table 2, it is a correspondence table of the pre-processed total amount of transaction invoices from the seller to the buyer, the pre-processed number of transactions from the seller to the buyer, the normalized pre-processed total amount of transaction invoices from the seller to the buyer, the normalized pre-processed number of transactions from the seller to the buyer, and the weights of the directed edges between the nodes corresponding to the seller and the nodes corresponding to the buyer for each pair of enterprises with a purchase and sale relationship in the largest connected subgraph.
[0152] Table 2
[0153]
[0154]
[0155] It should be noted that the weight calculated according to the above weight calculation formula is the weight of the edge in the direction from the node corresponding to the seller with a purchase and sale relationship to the node corresponding to the buyer in the maximum connected subgraph. In the embodiment of the present invention, the weight of the directed edge from the node corresponding to the seller to the node corresponding to the buyer in any two enterprises without a purchase and sale relationship is set to 0. For example, the weight of the edge from the node corresponding to enterprise A to the node corresponding to enterprise A (denoted as ω 11 ) is 0, the weight of the edge from the node corresponding to enterprise A to the node corresponding to enterprise D (denoted as ω 12 ) is 0, the weight of the edge from the node corresponding to enterprise A to the node corresponding to enterprise E (denoted as ω 15 ) is 0, ..., the weight of the edge from the node corresponding to enterprise B to the node corresponding to enterprise A (denoted as ω 21 ) is 0...etc., which will not be elaborated here.
[0156] S13. Determine the propagation probability between every two nodes in the maximum connected subgraph according to the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party.
[0157] In a specific implementation, for each pair of sellers and buyers in the maximum connected subgraph, the propagation probability between every two nodes in the maximum connected subgraph is determined by the following formula:
[0158]
[0159] Among them, T ij (i.e., T(j→i)) represents the probability of propagation from node j to node i, that is, the probability of the label of node j propagating to the label of node i;
[0160] ω ij Represents the weight of the directed edge from node i to node j;
[0161] n is the total number of nodes in the largest connected subgraph;
[0162] Represents the sum of the weights of the directed edges from node i to each node in the maximum connected subgraph.
[0163] Continuing with the above example, as shown in Table 3, it is a propagation probability table between nodes corresponding to every two enterprises in the maximum connected subgraph:
[0164] Table 3
[0165]
[0166]
[0167] Then the propagation probability matrix composed of the propagation probabilities between the nodes corresponding to each two enterprises in the above maximum connected subgraph is T = (T ij ), i=1~11, j=1~11.
[0168] S14. Based on the label marked for the designated tax-related abnormal enterprise, the labels marked for several other designated enterprises, and the propagation probability between every two nodes, predict the label of the unlabeled enterprise, where the label is the probability that the enterprise is a tax-related abnormal enterprise.
[0169] In specific implementation, the computer device predicts the label of an unlabeled enterprise based on the label marked for the designated tax-related abnormal enterprise, the labels marked for several other designated enterprises, and the propagation probability between every two nodes of the maximum connected subgraph, where the label is the probability that the enterprise is a tax-related abnormal enterprise.
[0170] Specifically, the labels of unlabeled enterprises can be predicted by following the steps below:
[0171] Step 1: Define the matrix Y∈(l+u)*C={Y i,C}, where l is the number of labeled enterprises, u is the number of unlabeled enterprises, C represents the category matrix to which the enterprise belongs, C = {1,0}, when C = 1, it means that the category of the enterprise is a tax-related abnormal enterprise, when C = 0, it means that the category of the enterprise is a tax-related credit enterprise, Y i,C Indicates the probability that node i is of category C, l<<u. In the embodiment of the present invention, the sum of the probability that the same enterprise is an abnormal tax-related enterprise and the probability that it is a credit tax-related enterprise is 1.
[0172] In specific implementation, the labels marked for the designated tax-related abnormal enterprises and the labels marked for several other designated enterprises are used as marking data, and the labels of the unlabeled enterprises are initialized and set to set values.
[0173] Continuing with the above example, if we know that enterprise A in the largest connected subgraph is an abnormal tax-related enterprise and enterprise H is an abnormal tax-related enterprise, then we label enterprise A with: Y 1,C =P(C=1)=1,Y 1,C =P(C=0)=0, that is, the probability that enterprise A is an abnormal tax-related enterprise is 1, and the probability that enterprise A is a credit-related tax-related enterprise is 0. Label enterprise H: Y 1,C =P(C=1)=0,Y 1,C =P(C=0)=1, that is, the probability that H enterprise is a tax-related abnormal enterprise is 0, and the probability that A enterprise is a tax-related credit enterprise is 1. According to the labels of the labeled enterprises A and B as input, the labels of B, C, D, E, F, G, I, J and K are predicted, that is, the probability that B, C, D, E, F, G, I, J and K are tax-related abnormal enterprises and the probability that they are tax-related credit enterprises are predicted respectively. Since the labeled enterprises are A and H, and the number of unlabeled enterprises is 9, the value of l in the defined matrix Y is 2, the value of u is 9, and the labels of the unlabeled enterprises are initialized, and the labels of the unlabeled enterprises are set to 0.5, that is, the probability that each unlabeled enterprise is a tax-related abnormal enterprise and the probability that it is a tax-related credit enterprise can be set to 0.5 respectively. In this way, the matrix Y is composed of the labeled values in Table 4, and the i-th row of the matrix Y represents the labeled probability of node i:
[0174] Table 4
[0175] node <![CDATA[Y i,C :P(C=1)]]> <![CDATA[Y i,C :P(C=0)]]> Company A 1 0 Company B 0.5 0.5 C Enterprise 0.5 0.5 D Enterprise 0.5 0.5 E-Enterprise 0.5 0.5 F Enterprise 0.5 0.5 G Enterprise 0.5 0.5 H Enterprise 0 1 I Enterprise 0.5 0.5 J Enterprise 0.5 0.5 K Enterprise 0.5 0.5
[0176] That is, the matrix Y is:
[0177] Step 2: construct a propagation probability matrix according to the propagation probability between every two nodes of the maximum connected subgraph.
[0178] In the specific implementation, the above example is still continued, and the propagation probability matrix T composed of the propagation probabilities between the nodes corresponding to every two enterprises in the above-mentioned maximum connected subgraph is the matrix T composed of the elements in Table 3.
[0179] Then, loop steps 3 and 4 to perform label propagation until convergence:
[0180] Step 3: Each node adds the weighted values of the annotations propagated by the surrounding nodes according to the propagation probability from the surrounding nodes to the node, and updates it to its own probability distribution to obtain Y t =T×Y t-1 , where t≥1, Y t represents the matrix Y after the tth update, Y t-1Represents the matrix Y before the t-th update.
[0181] In the specific implementation, each node adds the label values propagated by the surrounding nodes according to the propagation probability of its surrounding nodes to the node according to the weight of the edge between each node and its surrounding nodes, and updates it to its own probability distribution to obtain Y t =T×Y t-1 , where t≥1, Y t represents the matrix Y after the tth update, Y t-1 represents the matrix Y before the tth update, where, when t=1, Y t-1 is the initial setting matrix Y.
[0182] In specific implementation, the matrix Y in step one is multiplied by the propagation probability matrix T in step two to update the matrix Y: Y←T×Y.
[0183] Step 4: Reset the labels of the labeled enterprises in the matrix Y, limit the labeled data, and reassign the probability distribution of the labeled data to the initial value.
[0184] Repeat steps 3 and 4 until Y converges.
[0185] Label propagation algorithm description:
[0186] In the label propagation algorithm, (x1,y1),...,(x l ,y l ) indicates the labeled data;
[0187] Y L =(y1,……,y l )∈(1,…,C) represents the category of the labeled data, C is known and exists in the labeled data;
[0188] (x l+1 ,y l+1 ),……,(x l+u ,y l+u ) represents unlabeled data;
[0189] Y U ={y l+1 ,……,y l+u}: No label, satisfying l<<u, that is, the number of labeled data is much smaller than the number of unlabeled data;
[0190] X={x1,……,x l+u}∈R D ;
[0191] The problem described by the label propagation algorithm is: from X and Y L To predict Y U .
[0192] In the specific implementation, the labels of the marked enterprises in the updated matrix Y, i.e., the labels of enterprise A and enterprise H, are set to the initial values, i.e., the row corresponding to enterprise A (i.e., row 1) in the updated matrix Y is reassigned to: 1, 0, and the row corresponding to enterprise H (i.e., row 8) is assigned to 0, 1. Then, step 3 is executed until convergence. The labels of each node after convergence are shown in Table 5:
[0193] Table 5
[0194] node <![CDATA[Y i,C :P(C=1)]]> <![CDATA[Y i,C :P(C=0)]]> Company A 1 0 Company B 0 1 C Enterprise 0.43 0.57 D Enterprise 0.956 0.044 E-Enterprise 0.401 0.599 F Enterprise 0 1 G Enterprise 0 1 H Enterprise 0 1 I Enterprise 0 1 J Enterprise 0 1 K Enterprise 0 1
[0195] According to Table 5, among the unlabeled enterprises, the probability of Enterprise B, Enterprise F, Enterprise G, Enterprise H, Enterprise I, Enterprise J and Enterprise K being tax-related abnormal enterprises is 0, the probability of Enterprise C being a tax-related abnormal enterprise is 0.43, the probability of Enterprise D being a tax-related abnormal enterprise is 0.956, and the probability of Enterprise E being a tax-related abnormal enterprise is 0.401.
[0196] In an embodiment of the present invention, the weight of the directed edge from the starting node corresponding to the seller to the ending node corresponding to the buyer is calculated by using the total amount of transaction invoices from the seller to the buyer and the number of transactions. When the label propagation algorithm is used to predict the probability that an unlabeled enterprise is an abnormal tax-related enterprise, the weight of the directed edge calculated in the above manner is used to calculate the propagation probability between each two nodes. Compared with the prior art, the above method of calculating the edge weight makes the calculation of the closeness of the relationship between enterprises more accurate, and further, in the process of label propagation, the accuracy of predicting unlabeled labels is higher, thereby improving the accuracy of identifying tax-related risk enterprises.
[0197] S15. Identify whether the unlabeled enterprise is a tax risk enterprise based on the predicted label of the unlabeled enterprise.
[0198] In specific implementation, Figure 7 The steps shown in the figure can include the following steps to identify whether an unlabeled enterprise is a tax risk enterprise:
[0199] S41. When the probability that the predicted unlabeled enterprise is a tax-related abnormal enterprise is greater than or equal to a preset threshold, the unlabeled enterprise is determined to be a tax-related risk enterprise.
[0200] During specific implementation, the preset threshold may be set according to actual conditions. For example, it may be, but is not limited to, set to 0.6, which is not limited in the embodiment of the present invention.
[0201] S42: When the probability that the predicted unlabeled enterprise is a tax-related abnormal enterprise is less than the preset threshold, the unlabeled enterprise is determined to be a tax-related credit enterprise.
[0202] Assuming the preset threshold is 0.6, in the above example, Company D is a tax-risk company.
[0203] In the tax risk enterprise identification method provided in the embodiment of the present invention, a computer device responds to a trigger request for querying tax risk enterprises, obtains a maximum connected subgraph of a specified tax-related abnormal enterprise from a stored enterprise association graph, the enterprise association graph is a directed association graph constructed based on the transaction relationship between enterprises, with the first transaction party as the starting node, the second transaction party as the ending node, and the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party as directed edges, the maximum connected subgraph represents the maximum set that has a connected relationship with the specified tax-related abnormal enterprise, for each pair of the first transaction party and the second transaction party in the maximum connected subgraph, according to the total amount of transaction invoices from the first transaction party to the second transaction party in the pair of the first transaction party and the second transaction party The amount and number of transactions are used to determine the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in the pair of the first transaction party and the second transaction party, wherein the weight of the directed edge represents the closeness of the relationship between the first transaction party and the second transaction party. Furthermore, according to the determined weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party, the propagation probability between every two nodes in the maximum connected subgraph is determined. According to the labels marked for the designated tax-related abnormal enterprises, the labels marked for several other designated enterprises, and the propagation probability between every two nodes, the labels of the enterprises without labels are predicted, wherein the label is the probability that the enterprise is a tax-related abnormal enterprise. Furthermore, according to the predicted The labels of unlabeled enterprises are measured to identify whether the unlabeled enterprises are tax-risk enterprises. Compared with the prior art, in the embodiment of the present invention, a directed enterprise association graph is constructed according to the transaction relationship between enterprises to record the association relationship between enterprises. By finding the largest connected subgraph of known tax-related abnormal enterprises in the enterprise association graph, other enterprises that have an association relationship with the known tax-related abnormal enterprises can be quickly found. Then, according to the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party in each pair of the first transaction party and the second transaction party in the largest connected subgraph, the weight of the directed edge between the starting node corresponding to the first transaction party in each pair of the first transaction party and the second transaction party and the ending node corresponding to the second transaction party is determined as a representation The method measures the closeness of the relationship between the first transaction party and the second transaction party, and determines the propagation probability between each two nodes in the maximum connected subgraph according to the weight of the edge between the starting node corresponding to the first transaction party and the ending node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party. Based on the propagation probability between each two nodes and the labels of the labeled enterprises, the labels of the unlabeled enterprises are predicted according to the label propagation algorithm, that is, the probability of predicting that the unlabeled enterprises are tax-related abnormal enterprises is predicted, so as to identify whether the unlabeled enterprises are tax-related risk enterprises. The algorithm has low complexity and good classification effect, which saves a lot of time for subsequent judgment based on manual experience and analysis on whether enterprises associated with known tax-related abnormal enterprises have tax-related risks.Improved the efficiency and accuracy of identifying whether enterprises related to known tax-related abnormal enterprises are tax-related risk enterprises.
[0204] Based on the same inventive concept, an embodiment of the present invention also provides a device for identifying tax-related risk enterprises. Since the principle of solving the problem by the above-mentioned device for identifying tax-related risk enterprises is similar to that of the method for identifying tax-related risk enterprises, the implementation of the above-mentioned device can refer to the implementation of the method, and the repeated parts will not be repeated.
[0205] like Figure 8 As shown, it is a schematic diagram of the structure of the tax risk enterprise identification device provided by an embodiment of the present invention, which may include:
[0206] An acquisition unit 51 is used to respond to a trigger request for querying tax-risk enterprises and acquire a maximum connected subgraph of a specified tax-risk abnormal enterprise from a stored enterprise association graph, wherein the enterprise association graph is a directed association graph constructed based on the transaction relationship between enterprises, with the first transaction party as a starting node, the second transaction party as an end node, and the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party as directed edges, wherein the maximum connected subgraph represents the maximum set having a connected relationship with the specified tax-risk abnormal enterprise;
[0207] A first determining unit 52 is configured to determine, for each pair of first transaction parties and second transaction parties in the maximum connected subgraph, a weight of a directed edge from a node corresponding to the first transaction party in the pair of first transaction parties to a node corresponding to the second transaction party in the pair of first transaction parties and second transaction parties according to a total amount of transaction invoices and a number of transactions from the first transaction party to the second transaction party in the pair of first transaction parties, wherein the weight of the directed edge represents a closeness of a relationship between the first transaction party and the second transaction party;
[0208] A second determining unit 53 is used to determine the propagation probability between each two nodes in the maximum connected subgraph according to the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party;
[0209] The label prediction unit 54 is used to predict the label of the enterprise without a label based on the label marked for the designated tax-related abnormal enterprise, the labels marked for other designated enterprises, and the propagation probability between every two nodes, wherein the label is the probability that the enterprise is a tax-related abnormal enterprise;
[0210] The identification unit 55 is used to identify whether the unlabeled enterprise is a tax risk enterprise according to the predicted label of the unlabeled enterprise.
[0211] Preferably, the first determination unit 52 is specifically used to pre-process the total amount of transaction invoices from the first transaction party to the second transaction party and the number of transactions for each pair of the first transaction party and the second transaction party in the maximum connected subgraph; normalize the pre-processed total amount of transaction invoices from the first transaction party to the second transaction party and the number of transactions after pre-processing; and determine the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party based on the normalized total amount of transaction invoices from the first transaction party to the second transaction party and the normalized number of transactions after pre-processing.
[0212] Preferably, the first determining unit 52 is specifically configured to pre-process the total amount of the transaction invoice from the first transaction party to the second transaction party by using the following formula:
[0213]
[0214] in, Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party after pre-processing;
[0215] f se Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party;
[0216] fpje means fpje se max(fpje) represents the maximum value of the total amount of transaction invoices generated by all transaction parties in the maximum connected subgraph; and
[0217] The number of transactions from the first transaction party to the second transaction party is preprocessed using the following formula:
[0218]
[0219] in, Indicates the number of transactions from the first transaction party to the second transaction party after preprocessing;
[0220] count se Indicates the number of transactions from the first transaction party to the second transaction party;
[0221] count means count se max(count) represents the maximum value of the total number of transactions generated by all transaction parties in the maximum connected subgraph.
[0222] Preferably, the first determining unit 52 is specifically configured to normalize the pre-processed total amount of the transaction invoice from the first transaction party to the second transaction party by using the following formula:
[0223]
[0224] in, represents the normalized pre-processed total amount of the transaction invoice from the first transaction party to the second transaction party; and
[0225] The number of transactions from the first transaction party to the second transaction party after preprocessing is normalized by the following formula:
[0226]
[0227] in, Indicates the normalized preprocessed number of transactions from the first transaction party to the second transaction party.
[0228] Preferably, the first determining unit 52 is specifically configured to determine the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party by using the following formula:
[0229]
[0230] Wherein, ω represents the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party.
[0231] Preferably, the second determining unit 53 is specifically configured to determine the propagation probability between every two nodes in the maximum connected subgraph by using the following formula:
[0232]
[0233] Among them, T ij represents the transmission probability from node j to node i;
[0234] ω ij Represents the weight of the directed edge from node i to node j;
[0235] n is the total number of nodes in the largest connected subgraph.
[0236] Preferably, the label prediction unit is specifically used to define a matrix Y∈(l+u)*C={Y i,C}, where l is the number of labeled enterprises, u is the number of unlabeled enterprises, C represents the category matrix to which the enterprise belongs, C = {1,0}, when C = 1, it represents an abnormal tax-related enterprise, when C = 0, it represents a tax-related credit enterprise, Y i,C represents the probability that node i belongs to category C;
[0237] A propagation probability matrix is formed according to the propagation probability between each two nodes; and label propagation is performed according to the following steps until convergence:
[0238] Each node adds the label values propagated by the surrounding nodes according to the propagation probability from its surrounding nodes to the node, and updates it to its own probability distribution to obtain Y t =T×Y t-1 , where t≥1, Y t represents the matrix Y after the tth update, Y t-1 Indicates the matrix Y before the t-th update;
[0239] Reset the labels of the labeled enterprises in the matrix Y, limit the labeled data, and reassign the probability distribution of the labeled data to the initial value.
[0240] Preferably, the identification unit 55 is specifically used to determine that the unlabeled enterprise is a tax-related risk enterprise when the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is greater than or equal to a preset threshold; and to determine that the unlabeled enterprise is a tax-related credit enterprise when the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is less than the preset threshold.
[0241] Based on the same technical concept, the embodiment of the present invention further provides a computer device 600, referring to Fig. 9 As shown, the computer device 600 is used to implement the tax risk enterprise identification method recorded in the above method embodiment. The computer device 600 of this embodiment may include: a memory 601, a processor 602, and a computer program stored in the memory and executable on the processor, such as a tax risk enterprise identification program. When the processor executes the computer program, the steps in the above tax risk enterprise identification method embodiments are implemented, such as Figure 2 Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented, for example, 51.
[0242] The specific connection medium between the memory 601 and the processor 602 is not limited in the embodiment of the present invention. Fig. 9 In the embodiment, the memory 601 and the processor 602 are connected via a bus 603. The bus 603 is connected to the processor 602 via a bus 603. Fig. 9 The connection between other components is shown by bold lines, which is only for schematic illustration and is not intended to be limiting. The bus 603 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig. 9 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.
[0243] The memory 601 may be a volatile memory, such as a random-access memory (RAM); the memory 601 may also be a non-volatile memory, such as a read-only memory, a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), or the memory 601 may be any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 601 may be a combination of the above memories.
[0244] Processor 602, used to implement Figure 2 A method for identifying tax risk enterprises shown in FIG. 1 includes:
[0245] The processor 602 is used to call the computer program stored in the memory 601 to execute the following Figure 2 In step S11, in response to a trigger request for querying tax-related risk enterprises, a maximum connected subgraph of a specified tax-related abnormal enterprise is obtained from a stored enterprise association graph, wherein the enterprise association graph is a directed association graph constructed based on the transaction relationship between enterprises, with the first transaction party as a starting node, the second transaction party as an end node, and the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party as directed edges, and step S12, for each pair of the first transaction party and the second transaction party in the maximum connected subgraph, according to the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party in the pair of the first transaction party and the second transaction party, the corresponding The weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party, step S13, determining the propagation probability between every two nodes in the maximum connected subgraph according to the determined weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party, step S14, predicting the label of the unlabeled enterprise according to the label marked for the designated tax-related abnormal enterprise, the label marked for other designated enterprises, and the propagation probability between every two nodes, the label is the probability that the enterprise is a tax-related abnormal enterprise, and step S15, identifying whether the unlabeled enterprise is a tax-related risk enterprise according to the predicted label of the unlabeled enterprise.
[0246] An embodiment of the present application also provides a computer-readable storage medium that stores computer-executable instructions required to execute the above-mentioned processor, which includes a program for executing the above-mentioned processor.
[0247] In some possible implementations, various aspects of the tax risk enterprise identification method provided by the present invention may also be implemented in the form of a program product, which includes a program code. When the program product is run on a computer device, the program code is used to enable the computer device to execute the steps of the tax risk enterprise identification method according to various exemplary embodiments of the present invention described above in this specification. For example, the computer device may execute the following steps: Figure 2 In step S11, in response to a trigger request for querying tax-related risk enterprises, a maximum connected subgraph of a specified tax-related abnormal enterprise is obtained from a stored enterprise association graph, wherein the enterprise association graph is a directed association graph constructed based on the transaction relationship between enterprises, with the first transaction party as a starting node, the second transaction party as an end node, and the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party as directed edges, and step S12, for each pair of the first transaction party and the second transaction party in the maximum connected subgraph, according to the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party in the pair of the first transaction party and the second transaction party, the corresponding The weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party, step S13, determining the propagation probability between every two nodes in the maximum connected subgraph according to the determined weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party, step S14, predicting the label of the unlabeled enterprise according to the label marked for the designated tax-related abnormal enterprise, the label marked for other designated enterprises, and the propagation probability between every two nodes, the label is the probability that the enterprise is a tax-related abnormal enterprise, and step S15, identifying whether the unlabeled enterprise is a tax-related risk enterprise according to the predicted label of the unlabeled enterprise.
[0248] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0249] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0250] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0251] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0252] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0253] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A method for identifying tax-related risk enterprises, characterized in that: include: In response to a trigger request for querying tax-risk enterprises, a maximum connected subgraph of a specified tax-risk abnormal enterprise is obtained from a stored enterprise association graph, wherein the enterprise association graph is a directed association graph constructed based on the transaction relationship between enterprises, with the first transaction party as a starting node, the second transaction party as an end node, and the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party as directed edges, and the maximum connected subgraph represents the maximum set having a connected relationship with the specified tax-risk abnormal enterprise; For each pair of the first transaction party and the second transaction party in the maximum connected subgraph, determine the weight of the directed edge from the node corresponding to the first transaction party in the pair of the first transaction party and the second transaction party to the node corresponding to the second transaction party according to the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party in the pair of the first transaction party and the second transaction party, wherein the weight of the directed edge represents the closeness of the relationship between the first transaction party and the second transaction party; Determining the propagation probability between every two nodes in the maximum connected subgraph according to the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party, specifically includes: determining the propagation probability between every two nodes in the maximum connected subgraph by the following formula: Among them, T ij represents the transmission probability from node j to node i; ω ij represents the weight of the directed edge from node i to node j; n is the total number of nodes in the maximum connected subgraph; According to the label marked for the designated tax-related abnormal enterprise, the labels marked for several other designated enterprises, and the propagation probability between every two nodes, predict the label of the unlabeled enterprise, wherein the label is the probability that the enterprise is a tax-related abnormal enterprise; According to the predicted label of the unlabeled enterprise, it is identified whether the unlabeled enterprise is a tax risk enterprise.
2. The method according to claim 1, characterized in that For each pair of the first transaction party and the second transaction party in the maximum connected subgraph, according to the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party in the pair of the first transaction party and the second transaction party, the weight of the directed edge from the node corresponding to the first transaction party in the pair of the first transaction party and the second transaction party to the node corresponding to the second transaction party is determined, specifically including: For each pair of the first transaction party and the second transaction party in the largest connected subgraph, pre-processing the total amount of transaction invoices from the first transaction party to the second transaction party and the number of transactions respectively; Normalizing the pre-processed total amount of the transaction invoices from the first transaction party to the second transaction party and the pre-processed number of transactions respectively; The weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party is determined according to the normalized preprocessed total amount of transaction invoices from the first transaction party to the second transaction party and the normalized preprocessed number of transactions.
3. The method according to claim 2, characterized in that Preprocessing the total amount of the transaction invoice from the first transaction party to the second transaction party includes: The total amount of the transaction invoice from the first transaction party to the second transaction party is pre-processed using the following formula: in, Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party after pre-processing; f se Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party; fpje means fpje se max(fpje) represents the maximum value of the total amount of transaction invoices generated by all transaction parties in the maximum connected subgraph; and The number of transactions from the first transaction party to the second transaction party is preprocessed using the following formula: in, Indicates the number of transactions from the first transaction party to the second transaction party after preprocessing; count se Indicates the number of transactions from the first transaction party to the second transaction party; count means count se max(count) represents the maximum value of the total number of transactions generated by all transaction parties in the maximum connected subgraph.
4. The method according to claim 3, characterized in that The pre-processed total amount of the transaction invoice from the first transaction party to the second transaction party is normalized, specifically including: The total amount of the pre-processed transaction invoice from the first transaction party to the second transaction party is normalized by the following formula: in, represents the normalized pre-processed total amount of the transaction invoice from the first transaction party to the second transaction party; and The number of transactions from the first transaction party to the second transaction party after preprocessing is normalized by the following formula: in, Indicates the normalized preprocessed number of transactions from the first transaction party to the second transaction party.
5. The method according to claim 4, characterized in that Determining the weight of a directed edge from a node corresponding to the first transaction party to a node corresponding to the second transaction party according to the normalized preprocessed total amount of transaction invoices from the first transaction party to the second transaction party and the normalized preprocessed number of transactions, specifically includes: The weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party is determined by the following formula: Wherein, ω represents the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party.
6. The method according to claim 1, characterized in that According to the labels marked for the designated tax-related abnormal enterprises, the labels marked for several other designated enterprises, and the propagation probability between every two nodes, the labels of the enterprises without labels are predicted, specifically including: Define the matrix Y∈(l+u)*C={Y i,C }, where l is the number of labeled enterprises, u is the number of unlabeled enterprises, C represents the category matrix to which the enterprise belongs, C = {1,0}, when C = 1, it represents an abnormal tax-related enterprise, when C = 0, it represents a tax-related credit enterprise, Y i,C represents the probability that node i belongs to category C; A propagation probability matrix is formed according to the propagation probability between each two nodes; and label propagation is performed according to the following steps until convergence: Each node adds the label values propagated by the surrounding nodes according to the propagation probability from its surrounding nodes to the node, and updates it to its own probability distribution to obtain Y t =T×Y t-1 , where t≥1, Y t represents the matrix Y after the tth update, Y t-1 Indicates the matrix Y before the t-th update; Reset the labels of the labeled enterprises in the matrix Y, limit the labeled data, and reassign the probability distribution of the labeled data to the initial value.
7. The method according to claim 1, characterized in that According to the predicted label of the unlabeled enterprise, identifying whether the unlabeled enterprise is a tax risk enterprise specifically includes: When it is determined that the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is greater than or equal to a preset threshold, the unlabeled enterprise is determined to be a tax-related risk enterprise; When it is determined that the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is less than the preset threshold, the unlabeled enterprise is determined to be a tax-related credit enterprise.
8. A device for identifying tax-related risky enterprises, characterized in that: include: an acquisition unit, configured to respond to a trigger request for querying tax-risk enterprises and acquire a maximum connected subgraph of a specified tax-related abnormal enterprise from a stored enterprise association graph, wherein the enterprise association graph is a directed association graph constructed based on a transaction relationship between enterprises, with the first transaction party as a starting node, the second transaction party as an end node, and the total amount of transaction invoices and the number of transactions from the first transaction party to the second transaction party as directed edges, wherein the maximum connected subgraph represents the maximum set having a connected relationship with the specified tax-related abnormal enterprise; A first determining unit is configured to determine, for each pair of first transaction parties and second transaction parties in the maximum connected subgraph, a weight of a directed edge from a node corresponding to the first transaction party in the pair of first transaction parties to a node corresponding to the second transaction party in the pair of first transaction parties and second transaction parties according to a total amount of transaction invoices and a number of transactions from the first transaction party to the second transaction party in the pair of first transaction parties, wherein the weight of the directed edge represents a closeness of a relationship between the first transaction party and the second transaction party; A second determining unit, configured to determine a propagation probability between each two nodes in the maximum connected subgraph according to the weight of a directed edge from a node corresponding to the first transaction party to a node corresponding to the second transaction party in each pair of the first transaction party and the second transaction party; The second determining unit is specifically configured to determine the propagation probability between every two nodes in the maximum connected subgraph by using the following formula: Among them, T ij represents the transmission probability from node j to node i; ω ij represents the weight of the directed edge from node i to node j; n is the total number of nodes in the maximum connected subgraph; A label prediction unit, used to predict the label of an unlabeled enterprise based on the label labeled for the designated tax-related abnormal enterprise, the labels labeled for several other designated enterprises, and the propagation probability between every two nodes, wherein the label is the probability that the enterprise is a tax-related abnormal enterprise; The identification unit is used to identify whether the unlabeled enterprise is a tax-related risk enterprise based on the predicted label of the unlabeled enterprise.
9. The device according to claim 8, characterized in that The first determining unit is specifically configured to pre-process the total amount of transaction invoices from the first transaction party to the second transaction party and the number of transactions for each pair of the first transaction party and the second transaction party in the maximum connected subgraph; Normalizing the pre-processed total amount of the transaction invoices from the first transaction party to the second transaction party and the pre-processed number of transactions respectively; The weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party is determined according to the normalized preprocessed total amount of transaction invoices from the first transaction party to the second transaction party and the normalized preprocessed number of transactions.
10. The device according to claim 9, characterized in that The first determining unit is specifically configured to pre-process the total amount of the transaction invoice from the first transaction party to the second transaction party using the following formula: in, Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party after pre-processing; f se Indicates the total amount of the transaction invoice from the first transaction party to the second transaction party; fpje means fpje se max(fpje) represents the maximum value of the total amount of transaction invoices generated by all transaction parties in the maximum connected subgraph; and The number of transactions from the first transaction party to the second transaction party is preprocessed using the following formula: in, Indicates the number of transactions from the first transaction party to the second transaction party after preprocessing; count se Indicates the number of transactions from the first transaction party to the second transaction party; count means count se max(count) represents the maximum value of the total number of transactions generated by all transaction parties in the maximum connected subgraph.
11. The device according to claim 10, characterized in that The first determining unit is specifically configured to normalize the total amount of the pre-processed transaction invoice from the first transaction party to the second transaction party using the following formula: in, represents the normalized pre-processed total amount of the transaction invoice from the first transaction party to the second transaction party; and The number of transactions from the first transaction party to the second transaction party after preprocessing is normalized by the following formula: in, Indicates the normalized preprocessed number of transactions from the first transaction party to the second transaction party.
12. The device according to claim 11, characterized in that The first determining unit is specifically configured to determine the weight of a directed edge from a node corresponding to the first transaction party to a node corresponding to the second transaction party by using the following formula: Wherein, ω represents the weight of the directed edge from the node corresponding to the first transaction party to the node corresponding to the second transaction party.
13. The device according to claim 8, characterized in that The label prediction unit is specifically used to define the matrix Y∈(l+u)*C={Y i,C }, where l is the number of labeled enterprises, u is the number of unlabeled enterprises, C represents the category matrix to which the enterprise belongs, C = {1,0}, when C = 1, it represents an abnormal tax-related enterprise, when C = 0, it represents a tax-related credit enterprise, Y i,C represents the probability that node i belongs to category C; A propagation probability matrix is formed according to the propagation probability between each two nodes; and label propagation is performed according to the following steps until convergence: Each node adds the label values propagated by the surrounding nodes according to the propagation probability from its surrounding nodes to the node, and updates it to its own probability distribution to obtain Y t =T×Y t-1 , where t≥1, Y t represents the matrix Y after the tth update, Y t-1 Indicates the matrix Y before the t-th update; Reset the labels of the labeled enterprises in the matrix Y, limit the labeled data, and reassign the probability distribution of the labeled data to the initial value.
14. The device according to claim 8, characterized in that The identification unit is specifically configured to determine that the unlabeled enterprise is a tax-related risk enterprise when the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is greater than or equal to a preset threshold; When it is determined that the predicted probability that the unlabeled enterprise is a tax-related abnormal enterprise is less than the preset threshold, the unlabeled enterprise is determined to be a tax-related credit enterprise.
15. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method for identifying tax-related risk enterprises as described in any one of claims 1 to 7 is implemented.
16. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for identifying tax-risk enterprises as described in any one of claims 1 to 7 are implemented.
