WO2023029758A1 - 一种企业经济犯罪侦查方法、系统及设备 - Google Patents

一种企业经济犯罪侦查方法、系统及设备 Download PDF

Info

Publication number
WO2023029758A1
WO2023029758A1 PCT/CN2022/105141 CN2022105141W WO2023029758A1 WO 2023029758 A1 WO2023029758 A1 WO 2023029758A1 CN 2022105141 W CN2022105141 W CN 2022105141W WO 2023029758 A1 WO2023029758 A1 WO 2023029758A1
Authority
WO
WIPO (PCT)
Prior art keywords
enterprise
node
economic
dimensional data
economic crime
Prior art date
Application number
PCT/CN2022/105141
Other languages
English (en)
French (fr)
Inventor
苏钰
陈卓
余晓填
王孝宇
Original Assignee
深圳云天励飞技术股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳云天励飞技术股份有限公司 filed Critical 深圳云天励飞技术股份有限公司
Publication of WO2023029758A1 publication Critical patent/WO2023029758A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06311—Scheduling, planning or task assignment for a person or group
    • G06Q10/063114—Status monitoring or status determination for a person or group
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/018—Certifying business or products
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/18—Legal services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services

Definitions

  • the present application relates to the technical field of information processing, and in particular to a method, system and equipment for investigating corporate economic crimes.
  • Economic investigation technology refers to the analysis of business operation information to speculate whether relevant companies are involved in economic crimes such as money laundering and illegal fund-raising, providing a forward-looking perspective for cracking economic crime cases, thereby protecting the property safety of the country and citizens.
  • the existing economic investigation technology mainly judges whether the target enterprise is a shell enterprise based on information such as registered address and social insurance purchase status.
  • the type of crime is single and the accuracy rate is low, which cannot meet the diversification of existing enterprise economic crime investigations. requirements and accuracy requirements.
  • the embodiment of the present application provides a method, system, and equipment for investigating enterprise economic crimes, which can effectively improve the efficiency and accuracy of enterprise economic crime investigations, and meet the diversified needs of existing enterprise economic crime investigations.
  • the present application provides a method for investigating corporate economic crimes, including: obtaining the first multi-dimensional data of the target enterprise; based on a pre-built point model, determining the points corresponding to each dimension data in the first multi-dimensional data item, and calculate the suspicious score of the target enterprise through the determined integral item, and the suspicious score is used to determine whether the target enterprise is involved in an economic crime; when the target enterprise is involved in an economic crime, obtain the target enterprise’s Second multi-dimensional data; according to the first multi-dimensional data and the second multi-dimensional data, determine the economic crime type of the target enterprise.
  • the suspicious score of the target enterprise is calculated by determining the integral item corresponding to the first dimension data through the integral model, thereby screening out the target enterprise suspected of economic crime, reducing the objects of enterprise economic crime investigation, and improving the investigation of enterprise economic crime
  • the efficiency, and further confirm and classify the target enterprises suspected of economic crimes through the second dimension data improve the accuracy of the investigation of corporate economic crimes, and subdivide the types of corporate economic crimes, satisfying the existing corporate economic crimes The diverse needs of investigation.
  • the determining the economic crime type of the target enterprise according to the first multidimensional data and the second multidimensional data includes:
  • the node label of the corresponding node is determined according to the node prediction value, and the economic crime type of the target enterprise is determined according to the node label.
  • the heterogeneous graph-based node classification model through the node information and the edge information,
  • the present application provides an enterprise economic crime investigation system, including:
  • the first multi-dimensional data acquisition unit is used to acquire the first multi-dimensional data of the target enterprise
  • An integral item determination unit configured to determine an integral item corresponding to each dimension data in the first multi-dimensional data
  • a suspicious score calculation unit configured to determine the score item corresponding to each dimension data in the first multi-dimensional data based on the pre-built score model, and calculate the suspicious score of the target enterprise through the determined score item, the said Suspiciousness scores are used to determine whether said targeted business is involved in economic crimes;
  • the second multi-dimensional data acquisition unit is used to obtain the second multi-dimensional data of the target enterprise when the target enterprise is involved in an economic crime;
  • the economic crime classification determining unit is configured to determine the economic crime type of the target enterprise according to the first multi-dimensional data and the second multi-dimensional data.
  • the unit for determining the classification of economic crimes includes:
  • a heterogeneous graph generating subunit configured to reconstruct the first multidimensional data and the second multidimensional data to generate a heterogeneous graph of the target enterprise, the heterogeneous graph including node information and edges information;
  • the node prediction value determination subunit is used to determine the node prediction value corresponding to the edge relationship between each node based on the heterogeneous graph node classification model through the node information and the edge information;
  • the economic crime classification determination subunit is used to determine the node label of the corresponding node according to the node prediction value, and determine the economic crime type of the target enterprise according to the node label
  • the present application provides an enterprise economic crime investigation device, including a processor, a memory, and a computer program stored in the memory and operable on the processor.
  • the processor executes the computer program, Implement the method described in the first aspect or any optional manner of the first aspect.
  • the present application provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the first aspect or any optional method of the first aspect the method described.
  • the embodiment of the present application provides a computer program product.
  • the computer program product runs on the enterprise economic crime investigation equipment
  • the enterprise economic crime investigation equipment executes the enterprise economic crime investigation method described in the first aspect above. step.
  • Fig. 1 is a schematic flow chart of a method for investigating corporate economic crimes provided by the embodiment of the present application
  • Fig. 2 is a schematic flow chart of a method for determining the type of economic crime of a target enterprise provided by an embodiment of the present application;
  • Fig. 3 is a schematic diagram of an image of a heterogeneous image provided by the embodiment of the present application.
  • Fig. 4 is a schematic structural diagram of an enterprise economic crime investigation system provided by an embodiment of the present application.
  • Fig. 5 is a schematic structural diagram of an enterprise economic crime investigation device provided by an embodiment of the present application.
  • references to “one embodiment” or “some embodiments” or the like described in the specification of the present application mean that a particular feature, structure or characteristic described in connection with the embodiment is included in one or more embodiments of the present application .
  • appearances of the phrases “in one embodiment,” “in some embodiments,” “in other embodiments,” “in other embodiments,” etc. in various places in this specification are not necessarily All refer to the same embodiment, but mean “one or more but not all embodiments” unless specifically stated otherwise.
  • the terms “including”, “comprising”, “having” and variations thereof mean “including but not limited to”, unless specifically stated otherwise.
  • FIG. 1 is a schematic flow diagram of a method for investigating corporate economic crimes provided by the embodiment of the present application, which is described in detail as follows:
  • Step S101 acquiring the first multi-dimensional data of the target enterprise.
  • the first multi-dimensional data includes but is not limited to whether the enterprise contains persons with criminal record for economic crimes, the attendance rate of core personnel, the activity track of core personnel, public information of the enterprise, and whether the company’s personnel and colleagues are involved in economic crimes
  • each dimension of data corresponds to at least one integral item.
  • the integral items corresponding to each dimension data can pass business logic analysis and data set testing. Obtained by means of debugging and inspection.
  • the first dimension data is enterprise-related data that does not involve deep privacy or sensitive information, and can be obtained from enterprise public information, enterprise-related information captured by web crawlers, and other data.
  • the target company is one or more companies that need to be investigated for economic crimes.
  • Step S102 based on the pre-built integral model, determine the integral item corresponding to each dimension data in the above-mentioned first multi-dimensional data, and calculate the suspicious score of the above-mentioned target enterprise through the determined integral item, and the above-mentioned suspicious score is used to determine the above-mentioned target Whether the enterprise is involved in economic crimes.
  • whether an enterprise contains persons with previous convictions for economic crimes is to determine whether the personnel of the target enterprise to be investigated include persons with previous convictions for economic crimes.
  • the face recognition of the target enterprise personnel can be carried out through the image recognition method, and matched with the internal criminal files of the public security to determine whether the target enterprise personnel to be investigated include persons with criminal records of economic crimes.
  • the corresponding integral item is A; when the enterprise does not contain persons with economic crime history, the corresponding integral item is B, and A and B are different values. In general, the value of A is greater than the value of B, and B can be zero.
  • the attendance rate of the core personnel is to determine the attendance rate of the core personnel of the target enterprise to be investigated, such as the management and personnel above the management level.
  • the attendance rate of core personnel can be counted through company attendance records, or the face recognition of target enterprise personnel can be performed through image recognition methods, and whether the core personnel is on duty can be recorded to calculate the attendance rate of the core personnel.
  • the scope of the attendance rate can be determined, and then the corresponding integral item can be determined. For example, the integral item corresponding to the attendance rate (0-15%) is C, and the attendance rate (15%- 30%), the corresponding integral item is D, the lower the attendance rate, the higher the value of the corresponding integral item.
  • the activity trajectory of core personnel is to determine the activity trajectory of the core personnel of the target enterprise to be investigated, such as the management and above, and determine whether the core personnel have approached or entered suspicious places.
  • Suspicious venues include, but are not limited to: venues suspected of gambling, enterprises included in the dishonest list, and enterprises with persons with prior economic crimes.
  • the activity tracks of core personnel can be obtained through city-level cameras.
  • corporate public information includes but is not limited to public violation information, public opinion risk information, and economic crime-related information automatically detected by natural language processing (NLP, Natural Language Processing) technology, such as whether corporate promotional materials contain financial fraud. Words, such as capital preservation and high profits.
  • NLP Natural Language Processing
  • the public information of the enterprise contains any one or more of the above-mentioned information related to economic crimes, the corresponding point item is G, or the corresponding point items are determined according to different information types; when the public information of the enterprise does not contain the above-mentioned information related to economic crimes , the corresponding integral term is H, and H can be zero.
  • whether the company personnel's companions include persons with previous economic crimes is to determine whether other persons traveling with the target enterprise personnel include persons with previous economic crimes.
  • Face recognition can be carried out on the companions of enterprise personnel through image recognition methods, and matched with the internal criminal files of the public security to determine whether the associates of enterprise personnel to be investigated include persons with criminal records for economic crimes.
  • the corresponding integral item is I; when the enterprise personnel's companions do not include economic crime criminals, the corresponding integral item is J, and I and J are different values.
  • the value of I is greater than the value of J, which can be zero.
  • the value of the corresponding integral item is greater.
  • the attendance rate of core personnel is a necessary condition for maintaining the normal operation of the company, and the attendance rate will be marked as high suspicious.
  • the integral item corresponding to the attendance rate (0-15%) is recorded as 60 points
  • the integral item corresponding to the attendance rate (15%-30%) is recorded as 50 points.
  • the existence of public opinion risk in the company's public information is not necessarily related to whether the company is suspected of economic crimes, so the impact of this item on the suspicious score will be small.
  • corresponding scoring rules are set for each dimension of data, such as those corresponding to enterprises whose core personnel include persons with criminal record for economic crimes Integral items are recorded as 60 points; core personnel's attendance rate is 0-15% corresponding to the integral item is recorded as 60 points, the attendance rate is 15%-30% corresponding to the integral item is recorded as 50 points, and the attendance rate is 30%-50%
  • the corresponding point item is recorded as 20 points; the point item corresponding to the enterprise that has approached or entered a suspicious place in the activity track of the core personnel is recorded as 20 points;
  • the score item is recorded as 50 points; the point item corresponding to the enterprise that includes persons with economic crime history among the company personnel's companions is recorded as 50 points; Enterprise) whose contact times exceed the predetermined number of times, for example, 15 times, the corresponding integral item is 30 points and other integral rules, and the integral model is constructed.
  • the integral model After building the integral model, input the obtained first dimension data into the integral model to determine the integral item corresponding to each dimension data, and then calculate the target enterprise’s value through the calculation formula of suspicious score provided by the integral model Questionable score.
  • the suspicious score calculation formula is used to add the points corresponding to each dimension data in the first multi-dimensional data, and the added value is the suspicious score of the target enterprise.
  • the suspicious score of the target enterprise is calculated according to the determined integral item and suspicious score calculation formula, wherein the above suspicious score calculation formula is:
  • y represents a suspicious score
  • x i represents the i-th dimension data in the above-mentioned first multidimensional data
  • n represents the dimension of the above-mentioned first multidimensional data, i ⁇ (1,n), n>1, n is an integer
  • ⁇ i represents the nonlinear parameter corresponding to x i , Represents the integral term of the i-th dimension data.
  • xi represents the attendance rate of the core personnel of the enterprise
  • ⁇ i represents the nonlinear parameter corresponding to the attendance rate of the core personnel of the enterprise, such as the attendance rate 0% ⁇ xi ⁇ 15%
  • Step S103 when the above-mentioned target company is involved in an economic crime, obtain the second multi-dimensional data of the above-mentioned target company.
  • the second multi-dimensional data includes, but is not limited to: whether there is undisclosed violation information and bank statement information.
  • the suspicious score of the target enterprise after the suspicious score of the target enterprise is calculated, it can be determined whether the target enterprise is involved in an economic crime according to the comparison result of the suspicious score and the predetermined score.
  • the predetermined score is the point used to determine whether the target enterprise is involved in economic crimes.
  • the predetermined score is 100 points. When the suspicious score of the target enterprise exceeds 100 points, the target enterprise is determined to be involved in economic crimes; More than 100 points, it is determined that the target enterprise is not involved in economic crimes.
  • Step S104 according to the above-mentioned first multi-dimensional data and the above-mentioned second multi-dimensional data, determine the type of economic crime of the above-mentioned target enterprise.
  • the economic crime classification result of the target enterprise can be quickly and accurately determined.
  • the economic crime classification results include that the target company is a shell company, the target company is an illegal fund-raising company, the target company is a financial fraud company, and so on.
  • To determine the economic crime classification result of the target enterprise is to determine whether the target enterprise is any one or more of a shell company, an illegal fund-raising company, or a financial fraud company.
  • Fig. 2 is a schematic flowchart of a method for determining the type of economic crime of a target enterprise provided by the embodiment of the present application
  • Fig. 3 is a schematic diagram of an image of a heterogeneous graph provided by the embodiment of the present application
  • the method for determining the economic crime type of the target enterprise provided in Figure 2 is described in detail below in conjunction with Figure 3, as follows:
  • Step S201 reconstructing the first multi-dimensional data and the second multi-dimensional data to generate a heterogeneous graph of the target enterprise, where the heterogeneous graph includes node information and edge information.
  • the node information is the information of enterprise nodes and personnel nodes.
  • each enterprise is independently called a node, and each enterprise node is accompanied by relevant enterprise characteristics, that is, data characteristics related to enterprises in the first multidimensional data and the second multidimensional data, such as target
  • the company's public information contains the characteristics of high cost poly.
  • set node labels, node labels are shell companies, illegal fund-raising companies and fraudulent companies, that is, the node labels correspond to the types of economic crimes of enterprises.
  • each enterprise personnel is independently called a node, and the personnel node is accompanied by related personal characteristics, that is, data characteristics related to individuals in the first multidimensional data and the second multidimensional data, such as the activities of core personnel Trajectories, criminal records, etc.
  • the edge information is information related to edges connecting companies to companies, edges connecting companies to individuals, and edges connecting individuals to individuals.
  • the edge connecting the enterprise and the enterprise represents the flow transaction between the two enterprises;
  • the edge connecting the enterprise and the individual represents the employment relationship between the enterprise and the individual;
  • the edge connecting the individual represents the There is a peer relationship between the two.
  • enterprise A is a company suspected of fraud, and its node label is a fraud company.
  • enterprise C and enterprise A There is a flow transaction between enterprise C and enterprise A; there is an employment relationship between Zhang San and Li Si and enterprise A; relation.
  • Enterprise B and enterprise C are enterprises with unknown labels. It is necessary to analyze the heterogeneous graph to carry out information fusion between nodes, find better information fusion results, and determine the node labels of enterprise B and enterprise C according to the information fusion results.
  • Type that is, to determine the type of economic crimes of enterprise B and enterprise C, that is, in Figure 3, enterprise B and enterprise C are both target enterprises, and their corresponding node labels need to be determined, so as to determine enterprise B and enterprise C according to the corresponding node labels C type of economic crime.
  • multiple economic abnormal indicators can be constructed through business logic, such as the number of low-value remittances in a single month, the total amount of remittances, etc.; for the flow transaction information between enterprises, it will According to the name of the counterparty enterprise, it is matched with the abnormal enterprise database to obtain the abnormal transaction data related to the enterprise.
  • Step S202 based on the heterogeneous graph node classification model, through the above node information and the above edge information, determine the node prediction value corresponding to the edge relationship between each node.
  • the heterogeneous graph node classification model is used to classify the nodes in the heterogeneous graph, realize the fusion of information between heterogeneous nodes, and determine the node prediction value of the corresponding node according to the fusion result.
  • the heterogeneous graph node classification model is a neural network model based on a relational graph attention neural network model, in which the fusion of information between heterogeneous graph nodes is realized, A more accurate node aggregation expression is thus found to enable accurate and efficient classification of nodes in heterogeneous graphs.
  • each enterprise node is accompanied by related enterprise characteristics, and the heterogeneous graph node classification model can use these enterprise characteristics to construct the characteristic matrix of enterprise nodes; each personnel node is accompanied by related personal characteristics, through the heterogeneous graph A node classification model can use these personal features to build a feature matrix of person nodes.
  • the edge information includes the edge relationship between each node.
  • the corresponding parameter matrix can be learned from the edge relationship between each node.
  • multi-head attention is introduced.
  • the force mechanism can obtain a more accurate node aggregation expression, so as to obtain the node prediction value corresponding to the edge relationship between each node.
  • the node prediction value corresponding to the edge relationship between each node is determined according to the above node information and the above edge information through the node aggregation expression determined by the node classification model of the heterogeneous graph.
  • the node aggregation expression is:
  • r represents the edge relationship in the above-mentioned heterogeneous graph
  • Y r represents the node prediction value of r in the above-mentioned heterogeneous graph
  • H represents the feature matrix of related nodes
  • W r represents r learned in the above-mentioned relational graph attention neural network The parameter matrix of .
  • Step S203 Determine the node label of the corresponding node according to the above-mentioned node prediction value, and determine the economic crime type of the above-mentioned target enterprise according to the above-mentioned node label.
  • the node label of the corresponding enterprise node can be determined according to the node prediction value, so that according to the node label, the economic value of the target enterprise can be determined. type of crime.
  • the suspicious score of the target enterprise is calculated by determining the integral item corresponding to the first dimension data through the integral model, thereby screening out the target enterprise suspected of economic crime, reducing the objects of enterprise economic crime investigation, and improving the efficiency of the enterprise.
  • the efficiency of economic crime investigation, and further confirm and classify the target enterprises suspected of economic crime through the second dimension data improve the accuracy of enterprise economic crime investigation, and subdivide the types of enterprise economic crime, satisfying the existing The diversified needs of corporate economic crime investigation.
  • the embodiments of the present application further provide device embodiments for realizing the above-mentioned method embodiments.
  • FIG. 4 is a schematic diagram of an enterprise economic crime investigation system provided by an embodiment of the present application. Each included unit is used to execute each step in the embodiment corresponding to FIG. 1 . For details, please refer to the relevant description in the embodiment corresponding to FIG. 1 . For ease of description, only the parts related to this embodiment are shown.
  • the enterprise economic crime detection system 4 includes:
  • the first multidimensional data acquisition unit 41 is used to acquire the first multidimensional data of the target enterprise
  • the suspicious score calculation unit 42 is configured to determine the integral item corresponding to each dimension data in the first multi-dimensional data based on the pre-built integral model, and calculate the suspicious score of the target enterprise through the determined integral item, and the suspicious score Used to determine whether the above-mentioned target enterprises are involved in economic crimes;
  • the second multi-dimensional data acquisition unit 43 is used to obtain the second multi-dimensional data of the above-mentioned target company when the above-mentioned target company is involved in an economic crime;
  • the economic crime classification determining unit 44 is configured to determine the economic crime type of the target enterprise according to the first multi-dimensional data and the second multi-dimensional data.
  • the suspicious score calculation unit 42 is specifically used for:
  • y represents a suspicious score
  • x i represents the i-th dimension data in the above-mentioned first multidimensional data
  • n represents the dimension of the above-mentioned first multidimensional data, i ⁇ (1,n), n>1, n is an integer
  • ⁇ i represents the nonlinear parameter corresponding to x i , Represents the integral term of the i-th dimension data.
  • the economic crime classification determination unit 44 includes:
  • the heterogeneous graph generating subunit is used to reconstruct the above-mentioned first multi-dimensional data and the above-mentioned second multi-dimensional data to generate the above-mentioned heterogeneous graph of the target enterprise, and the above-mentioned heterogeneous graph includes node information and edge information;
  • the node prediction value determination subunit is used to determine the node prediction value corresponding to the edge relationship between each node based on the heterogeneous graph node classification model through the above node information and the above edge information;
  • the economic crime classification determination subunit is used to determine the node label of the corresponding node according to the above-mentioned node prediction value, and determine the economic crime type of the above-mentioned target enterprise according to the above-mentioned node label
  • the economic crime classification determines subunits, which are specifically used for:
  • r represents the edge relationship in the above-mentioned heterogeneous graph
  • Y r represents the node prediction value of r in the above-mentioned heterogeneous graph
  • H represents the feature matrix of related nodes
  • W r represents r learned in the above-mentioned relational graph attention neural network The parameter matrix of .
  • the above-mentioned first multi-dimensional data includes: whether the enterprise includes persons with criminal convictions for economic crimes, the attendance rate of core personnel, the activity track of core personnel, the public information of the enterprise, and whether the company’s colleagues include persons with criminal convictions for economic crimes.
  • Each dimension of data corresponds to at least one Points item.
  • the above-mentioned second multi-dimensional data includes: whether there is undisclosed violation information and bank statement information.
  • the suspicious score of the target enterprise is calculated by determining the integral item corresponding to the first dimension data through the integral model, thereby screening out the target enterprise suspected of economic crime, reducing the objects of enterprise economic crime investigation, and improving the investigation of enterprise economic crime
  • the efficiency, and further confirm and classify the target enterprises suspected of economic crimes through the second dimension data improve the accuracy of the investigation of corporate economic crimes, and subdivide the types of corporate economic crimes, satisfying the existing corporate economic crimes The diverse needs of investigation.
  • Fig. 5 is a schematic diagram of an enterprise economic crime investigation device provided by an embodiment of the present application.
  • the enterprise economic crime detection device 5 of this embodiment includes: a processor 50 , a memory 51 and a computer program 52 stored in the memory 51 and operable on the processor 50 , such as a speech recognition program.
  • the processor 50 executes the computer program 52
  • the steps in the above-mentioned embodiments of the enterprise economic crime investigation method are implemented, for example, steps 101-104 shown in FIG. 1 .
  • the processor 50 executes the computer program 52
  • the functions of the modules/units in the above-mentioned device embodiments are implemented, for example, the functions of the units 41-44 shown in FIG. 4 .
  • the computer program 52 can be divided into one or more modules/units, and one or more modules/units are stored in the memory 51 and executed by the processor 50 to complete the present application.
  • One or more modules/units may be a series of computer program instruction segments capable of accomplishing specific functions, and the instruction segments are used to describe the execution process of the computer program 52 in the enterprise economic crime investigation device 5 .
  • the computer program 52 can be divided into a first multidimensional data acquisition unit 41, a suspicious score calculation unit 42, a second multidimensional data acquisition unit 43, and an economic crime classification determination unit 44.
  • a suspicious score calculation unit 42 for a suspicious score calculation unit 42
  • a second multidimensional data acquisition unit 43 for specific functions of each unit, please refer to the corresponding section in FIG. 1 Relevant descriptions in the embodiments are not repeated here.
  • the enterprise economic crime investigation device may include, but not limited to, a processor 50 and a memory 51 .
  • a processor 50 may include, but not limited to, a processor 50 and a memory 51 .
  • Fig. 5 is only an example of the enterprise economic crime investigation equipment 5, and does not constitute a limitation to the enterprise economic crime investigation equipment 5, and may include more or less components than those shown in the illustration, or combine some Components, or different components, such as enterprise economic crime detection equipment may also include input and output devices, network access devices, buses, and so on.
  • the so-called processor 50 can be a central processing unit (Central Processing Unit, CPU), and can also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.
  • the memory 51 may be an internal storage unit of the enterprise economic crime investigation device 5 , such as a hard disk or memory of the enterprise economic crime investigation device 5 .
  • Storer 51 also can be the external storage device of enterprise economic crime investigation equipment 5, such as the plug-in hard disk equipped on enterprise economic crime investigation equipment 5, smart memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card (Flash Card), etc.
  • the memory 51 may also include both an internal storage unit of the enterprise economic crime investigation device 5 and an external storage device.
  • the memory 51 is used to store computer programs and other programs and data required by the enterprise economic crime investigation equipment.
  • the memory 51 can also be used to temporarily store data that has been output or will be output.
  • the embodiment of the present application also provides a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the above-mentioned enterprise economic crime investigation method can be realized.
  • the embodiment of the present application provides a computer program product.
  • the computer program product runs on the enterprise economic crime investigation equipment, the enterprise economic crime investigation equipment can realize the above enterprise economic crime investigation method when executed.

Landscapes

  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Engineering & Computer Science (AREA)
  • Tourism & Hospitality (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Marketing (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Development Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Health & Medical Sciences (AREA)
  • Educational Administration (AREA)
  • Technology Law (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本申请提供一种企业经济犯罪侦查方法、系统及设备，涉及信息处理技术领域，能够有效地提高企业经济犯罪侦查的效率以及准确率，并满足现有企业经济犯罪侦查的多元化需求。该方法包括：获取目标企业的第一多维度数据；基于预构建的积分模型，确定所述第一多维度数据中每一维度数据对应的积分项，并通过所确定的积分项计算所述目标企业的可疑分数，所述可疑分数用于确定所述目标企业是否涉嫌经济犯罪；当所述目标企业涉嫌经济犯罪时，获得所述目标企业的第二多维度数据；根据所述第一多维度数据和所述第二多维度数据，确定所述目标企业的经济犯罪类型。

Description

一种企业经济犯罪侦查方法、系统及设备
本申请要求于2021年9月1日提交中国专利局，申请号为202111021441.6、发明名称为“一种企业经济犯罪侦查方法、系统及设备”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及信息处理技术领域，尤其涉及一种企业经济犯罪侦查方法、系统及设备。
背景技术
经济侦查技术，是指通过分析企业运作信息，推测相关企业是否涉嫌经济犯罪比如洗钱、非法集资等，为破获经济犯罪案件提供前瞻视角，从而保护国家和公民的财产安全。
随着网络经济的发展，经济犯罪活动日渐猖獗，经济侦查的重要性也愈发重要。然而，目前已有的经济侦查技术主要是根据注册地址、社保购买情况等信息判断目标企业是否为空壳企业，判断犯罪类型单一且准确率较低，无法满足现有企业经济犯罪侦查的多元化需求及准确率需求。
发明内容
本申请实施例提供了一种企业经济犯罪侦查方法、系统及设备，能够有效地提高企业经济犯罪侦查的效率以及准确率，并满足现有企业经济犯罪侦查的多元化需求。
第一方面，本申请提供一种企业经济犯罪侦查方法，包括：获取目标企业的第一多维度数据；基于预构建的积分模型，确定所述第一多维度数据中每一维度数据对应的积分项，并通过所确定的积分项计算所述目标企业的可疑分数，所述可疑分数用于确定所述目标企业是否涉嫌经济犯罪；当所述目标企业涉嫌经济犯罪时，获得所述目标企业的第二多维度数据；根据所述第 一多维度数据和所述第二多维度数据，确定所述目标企业的经济犯罪类型。
本申请实施例通过积分模型确定第一维度数据对应的积分项来计算目标企业的可疑分数，从而将涉嫌经济犯罪的目标企业筛选出来，减少了企业经济犯罪侦查的对象，提高了企业经济犯罪侦查的效率，并且通过第二维度数据进一步对涉嫌经济犯罪的目标企业进行确认及分类，提高了企业经济犯罪侦查的准确率，并且对企业经济犯罪的类型进行细分，满足了现有企业经济犯罪侦查的多元化需求。
示例性的，所述根据所述第一多维度数据和所述第二多维度数据，确定所述目标企业的经济犯罪类型，包括：
将所述第一多维度数据和所述第二多维度数据进行重构，生成所述目标企业的异构图，所述异构图中包括节点信息和边信息；
基于异构图节点分类模型，通过所述节点信息和所述边信息，确定各个节点之间的边关系对应的节点预测值；
根据所述节点预测值，确定对应节点的节点标签，并根据所述节点标签确定所述目标企业的经济犯罪类型。
示例性的，所述基于异构图节点分类模型，通过所述节点信息和所述边信息，
第二方面，本申请提供一种企业经济犯罪侦查系统，包括：
第一多维度数据获取单元，用于获取目标企业的第一多维度数据；
积分项确定单元，用于确定所述第一多维度数据中每一维度数据对应的积分项；
可疑分数计算单元，用于基于预构建的积分模型，确定所述第一多维度数据中每一维度数据对应的积分项，并通过所确定的积分项计算所述目标企业的可疑分数，所述可疑分数用于确定所述目标企业是否涉嫌经济犯罪；
第二多维度数据获取单元，用于当所述目标企业涉嫌经济犯罪时，获得所述目标企业的第二多维度数据；
经济犯罪分类确定单元，用于根据所述第一多维度数据和所述第二多维度数据，确定所述目标企业的经济犯罪类型。
具体的，经济犯罪分类确定单元，包括：
异构图生成子单元，用于将所述第一多维度数据和所述第二多维度数据 进行重构，生成所述目标企业的异构图，所述异构图中包括节点信息和边信息；
节点预测值确定子单元，用于基于异构图节点分类模型，通过所述节点信息和所述边信息，确定各个节点之间的边关系对应的节点预测值；
经济犯罪分类确定子单元，用于根据所述节点预测值，确定对应节点的节点标签，并根据所述节点标签确定所述目标企业的经济犯罪类型
第三方面，本申请提供一种企业经济犯罪侦查设备，包括处理器、存储器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如第一方面或第一方面的任意可选方式所述的方法。
第四方面，本申请提供一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现如第一方面或第一方面的任意可选方式所述的方法。
第五方面，本申请实施例提供了一种计算机程序产品，当计算机程序产品在企业经济犯罪侦查设备上运行时，使得企业经济犯罪侦查设备执行上述第一方面所述的企业经济犯罪侦查方法的步骤。
可以理解的是，上述第二方面至第五方面的有益效果可以参见上述第一方面中的相关描述，在此不再赘述。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是本申请实施例提供的一种企业经济犯罪侦查方法的流程示意图；
图2是本申请实施例提供的一种确定目标企业的经济犯罪类型的方法的流程示意图；
图3是本申请实施例提供的一种异构图的图像示意图；
图4是本申请实施例提供的一种企业经济犯罪侦查系统的结构示意图；
图5是本申请实施例提供的一种企业经济犯罪侦查设备的结构示意图。
具体实施方式
以下描述中，为了说明而不是为了限定，提出了诸如特定系统结构、技术之类的具体细节，以便透彻理解本申请实施例。然而，本领域的技术人员应当清楚，在没有这些具体细节的其它实施例中也可以实现本申请。在其它情况中，省略对众所周知的系统、装置、电路以及方法的详细说明，以免不必要的细节妨碍本申请的描述。
应当理解，在本申请说明书和所附权利要求书中使用的术语“和/或”是指相关联列出的项中的一个或多个的任何组合以及所有可能组合，并且包括这些组合。另外，在本申请说明书和所附权利要求书的描述中，术语“第一”、“第二”、“第三”等仅用于区分描述，而不能理解为指示或暗示相对重要性。
还应当理解，在本申请说明书中描述的参考“一个实施例”或“一些实施例”等意味着在本申请的一个或多个实施例中包括结合该实施例描述的特定特征、结构或特点。由此，在本说明书中的不同之处出现的语句“在一个实施例中”、“在一些实施例中”、“在其他一些实施例中”、“在另外一些实施例中”等不是必然都参考相同的实施例，而是意味着“一个或多个但不是所有的实施例”，除非是以其他方式另外特别强调。术语“包括”、“包含”、“具有”及它们的变形都意味着“包括但不限于”，除非是以其他方式另外特别强调。
在现有的企业经济犯罪侦查方法中，只使用单个维度数据比如地址、社保等进行经济侦查时，无法保证经济犯罪侦查的准确率，而且所得的犯罪类型单一。当结合个人行为和企业行为进行企业经济犯罪侦查是时，需要人工构造经济犯罪的行为特征，效率低而且无法保证经济犯罪侦查的准确率。而且由于数据的隐私性，无法获取到大量企业的隐私信息，基于假设可以直接获取海量公司的隐私信息的经济犯罪侦查方法，往往难以落地。
请参见图1，图1是本申请实施例提供的一种企业经济犯罪侦查方法的流程示意图，详述如下：
步骤S101，获取目标企业的第一多维度数据。
在本申请实施例中，为避免因使用如地址、社保等单个维度数据对企业进行经济犯罪侦查时导致的准确率低，而且犯罪类型单一的问题，通过获取 多维度数据，再根据所获取到的多维度数据对企业进行经济犯罪侦查，可以有效地提高企业经济犯罪侦查的准确率，而且能够更好地对企业的经济犯罪进行分类。
在本申请的一些实施例中，第一多维度数据包括但不限于企业是否包含经济犯罪前科人员、核心人员的出勤率、核心人员的活动轨迹、企业公开信息以及企业人员同行人是否包含经济犯罪前科人员，每一维度数据至少对应一个积分项。每一维度数据对应的积分项，可以通过业务逻辑分析、数据集测试。调试与检验等方式来获得。
还需要说明的是，第一维度数据为不涉及深度隐私或敏感信息的企业相关数据，可以从企业公开信息、网络爬虫抓取到的企业相关信息等数据中获取。
还需要说明的是，目标企业为需要进行经济犯罪侦查的一个或多个企业。
步骤S102，基于预构建的积分模型，确定上述第一多维度数据中每一维度数据对应的积分项，并通过所确定的积分项计算上述目标企业的可疑分数，上述可疑分数用于确定上述目标企业是否涉嫌经济犯罪。
在本申请实施例中，在确定第一多维度数据中每一维度数据对应的积分项后，将所确定的积分项输入预构建的积分模型中，计算目标企业的可疑分数，再根据积分模型输出的结果即计算得到的目标企业的可疑分数，确定目标企业是否涉嫌经济犯罪。
在实际应用中，企业是否包含经济犯罪前科人员，是确定待侦查的目标企业人员中是否包含有经济犯罪前科人员。可以通过图像识别方法对目标企业人员进行人脸识别，并与公安内部犯罪档案进行匹配以确定待侦查的目标企业人员中是否包含有经济犯罪前科人员。当企业包含经济犯罪前科人员时，对应积分项为A，当企业不包含经济犯罪前科人员时，对应的积分项为B，A与B为不同的数值。一般情况下，A的值大于B的值，B可以为零。
在实际应用中，核心人员的出勤率，是确定待侦查的目标企业的核心人员比如管理层及管理层以上的人员的出勤率。可以通过公司考勤记录来统计核心人员的出勤率，也可以通过图像识别方法对目标企业人员进行人脸识别，记录核心人员是否到岗，从而计算出该核心人员的出勤率。在确定核心人员的出勤率对应的积分项时，可以确定出勤率所属的范围，再确定对应的积分 项，比如出勤率(0～15％)对应的积分项为C，出勤率(15％～30％)对应的积分项为D，出勤率越低，对应的积分项的值越高。
在实际应用中，核心人员的活动轨迹，是确定待侦查的目标企业的核心人员比如管理层及管理层以上的人员的活动轨迹，确定核心人员是否接近或出入过可疑场所。可疑场所包括但不限于：被怀疑赌博的场所，被纳入失信名单的企业，有经济犯罪前科人员的企业。可以基于人脸识别方法，通过城市级摄像头获取核心人员的活动轨迹。当核心人员的活动轨迹为接近或出入过可疑场所时，对应的积分项为E，当核心人员的活动轨迹为未接近或出入过可疑场所时，对应的积分项为F，E的值高于F的值，F可以为零。
在实际应用中，企业公开信息包括但不限于公开的违规信息、舆论风险信息，通过自然语言处理(NLP，Natural Language Processing)技术自动检测的经济犯罪相关信息，比如企业宣传资料中是否包含金融诈骗字眼，比如保本高利等。当企业公开信息包含上述经济犯罪相关信息的任意一项或多项时，对应的积分项为G，或者依据不同的信息类型确定对应的积分项；当企业公开信息未包含上述经济犯罪相关信息时，对应的积分项为H，H可以为零。
在实际应用中，企业人员同行人是否包含经济犯罪前科人员，是确定与目标企业人员同行的其他人员是否包含经济犯罪前科人员。可以通过图像识别方法对企业人员同行人进行人脸识别，并与公安内部犯罪档案进行匹配以确定待侦查的企业人员同行人中是否包含有经济犯罪前科人员。当企业人员同行人中包含经济犯罪前科人员时，对应积分项为I，当企业人员同行人中不包含经济犯罪前科人员时，对应的积分项为J，I与J为不同的数值。I的值大于J的值，J可以为零。
需要说明的是，当某一维度数据被标记为高度可疑时，对应的积分项的值越大，比如核心人员的出勤率是维持公司正常运转的必要条件，出勤率高低将会被以高可疑度的形式标记，比如出勤率(0～15％)对应的积分项记为60分，出勤率(15％～30％)对应的积分项记为50分。而企业公开信息中存在舆论风险与企业是否涉嫌经济犯罪不一定相关，则该项对于可疑分数的影响也会较小。
在本申请的一些实施例中，在构建积分模型时，通过业务逻辑和数据分 析调试，对每一维度数据都设置有相应的积分规则，比如核心人员中包含有经济犯罪前科人员的企业对应的积分项记为60分；核心人员出勤率为0～15％对应的积分项记为60分，出勤率为15％～30％对应的积分项记为50分，出勤率为30％～50％对应的积分项记为20分；核心人员的活动轨迹中有接近或出入过可疑场所的企业对应的积分项记为20分；企业公开信息中出现涉嫌经济犯罪比如高本保利字眼的企业对应的积分项记为50分；企业人员同行人中包含有经济犯罪前科人员的企业对应的积分项记为50分；预定时间内比如一个月内与可疑企业(可疑分数高或有过经济犯罪历史的企业)的联系次数超过预定次数比如15次的企业对应的积分项为30分等积分规则，构建积分模型。
在构建好积分模型后，将所获取到的第一维度数据输入至积分模型中，可以确定每一维度数据对应的积分项，然后通过积分模型提供的可疑分数计算公式，可以计算得到目标企业的可疑分数。其中可疑分数计算公式用于将所述第一多维度数据中每一维度数据对应的积分项相加，相加所得的值即为所述目标企业的可疑分数。
在本申请实施例中，基于上述积分模型，根据所确定的积分项和可疑分数计算公式，计算上述目标企业的可疑分数，其中，上述可疑分数计算公式为：
Figure PCTCN2022105141-appb-000001
其中，y表示可疑分数；x i表示上述第一多维度数据中的第i维度数据；n表示上述第一多维度数据的维数，i∈(1，n)，n>1，n为整数；ω i表示x i对应的非线性参数，
Figure PCTCN2022105141-appb-000002
表示第i维度数据的积分项。
在一个具体的实施例中，x i表示企业核心人员的出勤率，ω i表示企业核心人员的出勤率对应的非线性参数，比如出勤率0％≤x i＜15％，则
Figure PCTCN2022105141-appb-000003
步骤S103，当上述目标企业涉嫌经济犯罪时，获得上述目标企业的第二多维度数据。
在本申请实施例中，第二多维度数据包括但不限于：是否存在未公开违规信息、银行流水信息。
在本申请的一些实施例中，计算得到目标企业的可疑分数后，可以根据可疑分数与预定分数的比较结果，确定目标企业是否涉嫌经济犯罪。其中预定分数是用于确定目标企业是否涉嫌经济犯罪的分数点，比如预定分数为100 分，当目标企业的可疑分数超过100分以上，则认定目标企业涉嫌经济犯罪；当目标企业的可疑分数不超过100分，则认定目标企业不涉嫌经济犯罪。
在实际应用中，确定目标企业是否涉嫌经济犯罪，可以为后续申请违法记录、银行流水等深层次涉及敏感隐私的数据即第二多维度数据提供准备或佐证，可以通过向公安机关、市监局、银行等机构提供目标企业涉嫌经济犯罪的佐证，从相关部门中获取涉嫌经济犯罪的目标企业的深层隐私信息，从而解决了无法获得企业敏感数据的问题，使得能够提高对企业经济犯罪类型进行准确分类，并且降低了需要获取的数据量，提高了经济犯罪侦查的效率。
步骤S104，根据上述第一多维度数据和上述第二多维度数据，确定上述目标企业的经济犯罪类型。
在本申请实施例中，通过结合第一多维度数据和第二多维度数据进行分析，可以快速且准确地确定目标企业的经济犯罪分类结果。
在本申请实施例中，经济犯罪分类结果包含目标企业为空壳公司、目标企业为非法集资公司、目标企业为金融诈骗公司等等。确定目标企业的经济犯罪分类结果即是确定目标企业是为空壳公司、非法集资公司还是金融诈骗公司中的任意一种或多种。
请参见图2和图3，图2是本申请实施例提供的一种确定目标企业的经济犯罪类型的方法的流程示意图，图3是本申请实施例提供的一种异构图的图像示意图；下面结合图3，对图2提供的确定目标企业的经济犯罪类型的方法进行详细描述，具体如下：
步骤S201，将上述第一多维度数据和上述第二多维度数据进行重构，生成上述目标企业的异构图，上述异构图中包括节点信息和边信息。
在本申请实施例中，节点信息为企业节点和人员节点的信息。对于企业节点来说，每一个企业独立称为一个节点，并且每个企业节点伴随着相关的企业特征，也即第一多维数据和第二多维度数据中与企业相关的数据特征，比如目标企业的公开信息中包含高本保利字眼的特征。对于已知的涉嫌经济犯罪的企业，设定节点标签，节点标签为空壳公司、非法集资公司和诈骗公司，即节点标签对应为企业的经济犯罪类型。对于人员节点来说，每个企业人员独立称为一个节点，人员节点伴随相关的个人特征，也即第一多维数据和第二多维度数据中与个人相关的数据特征，比如核心人员的活动轨迹、犯 罪记录等。
边信息为连接企业与企业之间的边、连接企业与个人之间的边、连接个人与个人之间的边的相关信息。其中连接企业与企业之间的边，代表两个企业之间存在流水交易行为；连接企业与个人之间的边，代表企业与个人之间存在雇佣关系；连接个人与个人之间的边，代表两者之间存在同行关系。
如图3所示，企业A为涉嫌诈骗的公司，其节点标签为诈骗公司。企业C与企业A之间存在流水交易行为；张三和李四与企业A之间为雇佣关系；王五与李四为在一定时期内的同行人，而王五与企业B之间为雇佣关系。企业B和企业C为未知标签的企业，需要通过分析该异构图以进行节点间的信息融合，找到更好的信息融合结果，根据信息融合结果确定企业B和企业C的节点标签为哪一类型，即确定企业B和企业C的经济犯罪类型，也即在图3中，企业B和企业C均为目标企业，需要确定其对应的节点标签，从而根据对应的节点标签确定企业B和企业C的经济犯罪类型。
在本申请的一些实施例中，对于企业的银行流水，通过业务逻辑可以构建多个经济异常指标，比如单月内低额汇款次数、总汇款额度等；对于企业之间的流水交易信息，会根据交易对手企业名称与异常企业库进行匹配，以得到企业相关的异常交易数据。
步骤S202，基于异构图节点分类模型，通过上述节点信息和上述边信息，确定各个节点之间的边关系对应的节点预测值。
在本申请实施例中，异构图节点分类模型用于对异构图中的节点进行分类，实现异构节点间信息的融合，从而根据融合结果确定对应节点的节点预测值。
在本申请的一些实施例中，异构图节点分类模型为基于关系型图注意力神经网络模型进行建模的一个神经网络模型，在该模型中，实现了异构图节点间信息的融合，从而找到更准确的节点聚合表达式，以使得能够对准确有效地对异构图中的节点进行分类。
在实际应用中，每个企业节点伴随着相关的企业特征，通过异构图节点分类模型可以使用这些企业特征构建企业节点的特征矩阵；每个人员节点伴随着相关的个人特征，通过异构图节点分类模型可以使用这些个人特征构建人员节点的特征矩阵。
在实际应用中，边信息中包括各个节点之间的边关系，通过异构图节点分类模型可以对各个节点之间的边关系学习得到对应的参数矩阵，在学习的过程中，引入了多头注意力机制，可以得到更准确的节点聚合表达，从而得到各个节点之间的边关系对应的节点预测值。
在本申请的一些实施例中，通过异构图节点分类模型确定的节点聚合表达式，根据上述节点信息和上述边信息，确定各个节点之间的边关系对应的节点预测值。
具体的，节点聚合表达式为：
Y r＝HW r
其中，r表示上述异构图中的边关系，Y r表示上述异构图中r的节点预测值，H表示相关节点的特征矩阵，W r表示r在上述关系型图注意力神经网络学习得到的参数矩阵。
步骤S203，根据上述节点预测值，确定对应节点的节点标签，并根据上述节点标签确定上述目标企业的经济犯罪类型。
在本申请实施例中，在得到各个节点之间的边关系对应的节点预测值后，可以根据该节点预测值确定对应的企业节点的节点标签，从而根据该节点标签，可以确定目标企业的经济犯罪类型。
在本申请实施例中，通过积分模型确定第一维度数据对应的积分项来计算目标企业的可疑分数，从而将涉嫌经济犯罪的目标企业筛选出来，减少了企业经济犯罪侦查的对象，提高了企业经济犯罪侦查的效率，并且通过第二维度数据进一步对涉嫌经济犯罪的目标企业进行确认及分类，提高了企业经济犯罪侦查的准确率，并且对企业经济犯罪的类型进行细分，满足了现有企业经济犯罪侦查的多元化需求。
应理解，上述实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
基于上述实施例所提供的企业经济犯罪侦查方法，本申请实施例进一步给出实现上述方法实施例的装置实施例。
请参见图4，图4是本申请实施例提供的企业经济犯罪侦查系统的示意图。包括的各单元用于执行图1对应的实施例中的各步骤。具体请参阅图1 对应的实施例中的相关描述。为了便于说明，仅示出了与本实施例相关的部分。参见图4，企业经济犯罪侦查系统4包括：
第一多维度数据获取单元41，用于获取目标企业的第一多维度数据；
可疑分数计算单元42，用于基于预构建的积分模型，确定上述第一多维度数据中每一维度数据对应的积分项，并通过所确定的积分项计算上述目标企业的可疑分数，上述可疑分数用于确定上述目标企业是否涉嫌经济犯罪；
第二多维度数据获取单元43，用于当上述目标企业涉嫌经济犯罪时，获得上述目标企业的第二多维度数据；
经济犯罪分类确定单元44，用于根据上述第一多维度数据和上述第二多维度数据，确定上述目标企业的经济犯罪类型。
具体的，可疑分数计算单元42，具体用于：
基于上述积分模型，根据所确定的积分项和可疑分数计算公式，计算上述目标企业的可疑分数，其中，上述可疑分数计算公式为：
Figure PCTCN2022105141-appb-000004
其中，y表示可疑分数；x i表示上述第一多维度数据中的第i维度数据；n表示上述第一多维度数据的维数，i∈(1，n)，n>1，n为整数；ω i表示x i对应的非线性参数，
Figure PCTCN2022105141-appb-000005
表示第i维度数据的积分项。
具体的，经济犯罪分类确定单元44，包括：
异构图生成子单元，用于将上述第一多维度数据和上述第二多维度数据进行重构，生成上述目标企业的异构图，上述异构图中包括节点信息和边信息；
节点预测值确定子单元，用于基于异构图节点分类模型，通过上述节点信息和上述边信息，确定各个节点之间的边关系对应的节点预测值；
经济犯罪分类确定子单元，用于根据上述节点预测值，确定对应节点的节点标签，并根据上述节点标签确定上述目标企业的经济犯罪类型
具体的，经济犯罪分类确定子单元，具体用于：
通过异构图节点分类模型确定的节点聚合表达式，根据上述节点信息和上述边信息，确定各个节点之间的边关系对应的节点预测值。
上述节点聚合表达式为：
Y r＝HW r
其中，r表示上述异构图中的边关系，Y r表示上述异构图中r的节点预测值，H表示相关节点的特征矩阵，W r表示r在上述关系型图注意力神经网络学习得到的参数矩阵。
上述第一多维度数据包括：企业是否包含经济犯罪前科人员、核心人员的出勤率、核心人员的活动轨迹、企业公开信息以及企业人员同行人是否包含经济犯罪前科人员，每一维度数据至少对应一个积分项。
上述第二多维度数据包括：是否存在未公开违规信息、银行流水信息。
本申请实施例通过积分模型确定第一维度数据对应的积分项来计算目标企业的可疑分数，从而将涉嫌经济犯罪的目标企业筛选出来，减少了企业经济犯罪侦查的对象，提高了企业经济犯罪侦查的效率，并且通过第二维度数据进一步对涉嫌经济犯罪的目标企业进行确认及分类，提高了企业经济犯罪侦查的准确率，并且对企业经济犯罪的类型进行细分，满足了现有企业经济犯罪侦查的多元化需求。
需要说明的是，上述模块之间的信息交互、执行过程等内容，由于与本申请方法实施例基于同一构思，其具体功能及带来的技术效果，具体可参见方法实施例部分，此处不再赘述。
图5是本申请实施例提供的企业经济犯罪侦查设备的示意图。如图5所示，该实施例的企业经济犯罪侦查设备5包括：处理器50、存储器51以及存储在存储器51中并可在处理器50上运行的计算机程序52，例如语音识别程序。处理器50执行计算机程序52时实现上述各个企业经济犯罪侦查方法实施例中的步骤，例如图1所示的步骤101-104。或者，处理器50执行计算机程序52时实现上述各装置实施例中各模块/单元的功能，例如图4所示单元41-44的功能。
示例性的，计算机程序52可以被分割成一个或多个模块/单元，一个或者多个模块/单元被存储在存储器51中，并由处理器50执行，以完成本申请。一个或多个模块/单元可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述计算机程序52在企业经济犯罪侦查设备5中的执行过程。例如，计算机程序52可以被分割成第一多维度数据获取单元41、可疑分数计算单元42、第二多维度数据获取单元43、经济犯罪分类确定单元44，各单元具体功能请参阅图1对应的实施例中地相关描述，此处不赘述。
企业经济犯罪侦查设备可包括，但不仅限于，处理器50、存储器51。本领域技术人员可以理解，图5仅仅是企业经济犯罪侦查设备5的示例，并不构成对企业经济犯罪侦查设备5的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如企业经济犯罪侦查设备还可以包括输入输出设备、网络接入设备、总线等。
所称处理器50可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
存储器51可以是企业经济犯罪侦查设备5的内部存储单元，例如企业经济犯罪侦查设备5的硬盘或内存。存储器51也可以是企业经济犯罪侦查设备5的外部存储设备，例如企业经济犯罪侦查设备5上配备的插接式硬盘，智能存储卡(Smart Media Card,SMC)，安全数字(Secure Digital,SD)卡，闪存卡(Flash Card)等。进一步地，存储器51还可以既包括企业经济犯罪侦查设备5的内部存储单元也包括外部存储设备。存储器51用于存储计算机程序以及企业经济犯罪侦查设备所需的其他程序和数据。存储器51还可以用于暂时地存储已经输出或者将要输出的数据。
本申请实施例还提供了一种计算机可读存储介质，计算机可读存储介质存储有计算机程序，计算机程序被处理器执行时可实现上述企业经济犯罪侦查方法。
本申请实施例提供了一种计算机程序产品，当计算机程序产品在企业经济犯罪侦查设备上运行时，使得企业经济犯罪侦查设备执行时实现可实现上述企业经济犯罪侦查方法。
所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，仅以上述各功能单元、模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能单元、模块完成，即将装置的内部结构划分成不同的功能单元或模块，以完成以上描述的全部或者部分功能。实施例中的各功能单元、模块可以集成在一个处理单元中，也可以是各个单元单独物理 存在，也可以两个或两个以上单元集成在一个单元中，上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。另外，各功能单元、模块的具体名称也只是为了便于相互区分，并不用于限制本申请的保护范围。上述系统中单元、模块的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述或记载的部分，可以参见其它实施例的相关描述。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
以上所述实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围，均应包含在本申请的保护范围之内。

Claims (10)

  1. 一种企业经济犯罪侦查方法，其特征在于，所述企业经济犯罪侦察方法包括：
    获取目标企业的第一多维度数据；
    基于预构建的积分模型，确定所述第一多维度数据中每一维度数据对应的积分项，并通过所确定的积分项计算所述目标企业的可疑分数，所述可疑分数用于确定所述目标企业是否涉嫌经济犯罪；
    当所述目标企业涉嫌经济犯罪时，获得所述目标企业的第二多维度数据；
    根据所述第一多维度数据和所述第二多维度数据，确定所述目标企业的经济犯罪类型。
  2. 如权利要求1所述的企业经济犯罪侦查方法，其特征在于，所述基于预构建的积分模型，确定所述第一多维度数据中每一维度数据对应的积分项，并通过所确定的积分项计算所述目标企业的可疑分数，包括：
    基于所述积分模型，根据所确定的积分项和可疑分数计算公式，计算所述目标企业的可疑分数，其中，所述可疑分数计算公式用于将所述第一多维度数据中每一维度数据对应的积分项相加，相加所得的值即为所述目标企业的可疑分数。
  3. 如权利要求1或2所述的企业经济犯罪侦查方法，其特征在于，所述根据所述第一多维度数据和所述第二多维度数据，确定所述目标企业的经济犯罪类型，包括：
    将所述第一多维度数据和所述第二多维度数据进行重构，生成所述目标企业的异构图，所述异构图中包括节点信息和边信息；
    基于异构图节点分类模型，通过所述节点信息和所述边信息，确定各个节点之间的边关系对应的节点预测值；
    根据所述节点预测值，确定对应节点的节点标签，并根据所述节点标签确定所述目标企业的经济犯罪类型。
  4. 如权利要求3所述的企业经济犯罪侦查方法，其特征在于，所述基于异构图节点分类模型，通过所述节点信息和所述边信息，确定各个节点之间的边关系对应的节点预测值，包括：
    通过异构图节点分类模型确定的节点聚合表达式，根据所述节点信息和 所述边信息，确定各个节点之间的边关系对应的节点预测值。
  5. 如权利要求1所述的企业经济犯罪侦查方法，其特征在于，所述第一多维度数据包括：企业是否包含经济犯罪前科人员、核心人员的出勤率、核心人员的活动轨迹、企业公开信息以及企业人员同行人是否包含经济犯罪前科人员，每一维度数据至少对应一个积分项。
  6. 如权利要求1所述的企业经济犯罪侦查方法，其特征在于，所述第二多维度数据包括：是否存在未公开违规信息、银行流水信息。
  7. 一种企业经济犯罪侦查系统，其特征在于，所述企业经济犯罪侦察系统包括：
    第一多维度数据获取单元，用于获取目标企业的第一多维度数据；
    积分项确定单元，用于确定所述第一多维度数据中每一维度数据对应的积分项；
    可疑分数计算单元，用于基于预构建的积分模型，确定所述第一多维度数据中每一维度数据对应的积分项，并通过所确定的积分项计算所述目标企业的可疑分数，所述可疑分数用于确定所述目标企业是否涉嫌经济犯罪；
    第二多维度数据获取单元，用于当所述目标企业涉嫌经济犯罪时，获得所述目标企业的第二多维度数据；
    经济犯罪分类确定单元，用于根据所述第一多维度数据和所述第二多维度数据，确定所述目标企业的经济犯罪类型。
  8. 如权利要求7所述的企业经济犯罪侦查系统，其特征在于，所述经济犯罪分类确定单元，包括：
    异构图生成子单元，用于将所述第一多维度数据和所述第二多维度数据进行重构，生成所述目标企业的异构图，所述异构图中包括节点信息和边信息；
    节点预测值确定子单元，用于基于异构图节点分类模型，通过所述节点信息和所述边信息，确定各个节点之间的边关系对应的节点预测值；
    经济犯罪分类确定子单元，用于根据所述节点预测值，确定对应节点的节点标签，并根据所述节点标签确定所述目标企业的经济犯罪类型。
  9. 一种企业经济犯罪侦查设备，其特征在于，包括处理器、存储器以及存储在所述存储器中并可在所述处理器上运行的计算机程序，其特征在于， 所述处理器执行所述计算机程序时实现如权利要求1至7任一项所述的企业经济犯罪侦查方法。
  10. 一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现如权利要求1至7任一项所述的企业经济犯罪侦查方法。
PCT/CN2022/105141 2021-09-01 2022-07-12 一种企业经济犯罪侦查方法、系统及设备 WO2023029758A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111021441.6A CN113919630A (zh) 2021-09-01 2021-09-01 一种企业经济犯罪侦查方法、系统及设备
CN202111021441.6 2021-09-01

Publications (1)

Publication Number Publication Date
WO2023029758A1 true WO2023029758A1 (zh) 2023-03-09

Family

ID=79233616

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/105141 WO2023029758A1 (zh) 2021-09-01 2022-07-12 一种企业经济犯罪侦查方法、系统及设备

Country Status (2)

Country Link
CN (1) CN113919630A (zh)
WO (1) WO2023029758A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113919630A (zh) * 2021-09-01 2022-01-11 深圳云天励飞技术股份有限公司 一种企业经济犯罪侦查方法、系统及设备

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170017887A1 (en) * 2015-07-17 2017-01-19 Wipro Limited Methods and systems for detecting financial crimes in an enterprise
CN109949154A (zh) * 2018-12-17 2019-06-28 深圳平安综合金融服务有限公司 客户信息分类方法、装置、计算机设备和存储介质
CN110689438A (zh) * 2019-08-26 2020-01-14 深圳壹账通智能科技有限公司 企业类金融风险评分方法、装置、计算机设备及存储介质
CN112037038A (zh) * 2020-09-02 2020-12-04 中国银行股份有限公司 银行信贷风险预测方法及装置
CN113919630A (zh) * 2021-09-01 2022-01-11 深圳云天励飞技术股份有限公司 一种企业经济犯罪侦查方法、系统及设备

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112990946B (zh) * 2021-03-31 2024-05-14 建信金融科技有限责任公司 一种企业违约预测方法、装置、介质及电子设备

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170017887A1 (en) * 2015-07-17 2017-01-19 Wipro Limited Methods and systems for detecting financial crimes in an enterprise
CN109949154A (zh) * 2018-12-17 2019-06-28 深圳平安综合金融服务有限公司 客户信息分类方法、装置、计算机设备和存储介质
CN110689438A (zh) * 2019-08-26 2020-01-14 深圳壹账通智能科技有限公司 企业类金融风险评分方法、装置、计算机设备及存储介质
CN112037038A (zh) * 2020-09-02 2020-12-04 中国银行股份有限公司 银行信贷风险预测方法及装置
CN113919630A (zh) * 2021-09-01 2022-01-11 深圳云天励飞技术股份有限公司 一种企业经济犯罪侦查方法、系统及设备

Also Published As

Publication number Publication date
CN113919630A (zh) 2022-01-11

Similar Documents

Publication Publication Date Title
US11810204B2 (en) Artificial intelligence transaction risk scoring and anomaly detection
Suzumura et al. Towards federated graph learning for collaborative financial crimes detection
De Roux et al. Tax fraud detection for under-reporting declarations using an unsupervised machine learning approach
US8458090B1 (en) Detecting fraudulent mobile money transactions
JP6715838B2 (ja) 文書内の潜在的重要事実を自動識別するシステム及び方法
Bologa et al. Big data and specific analysis methods for insurance fraud detection.
Mangala et al. Auditors’ perceptions of the effectiveness of fraud prevention and detection methods
KR20180060044A (ko) 클라우드 환경에서 개인정보 보호를 지원하는 p2p 중개 보안 시스템
Starnini et al. Smurf-based anti-money laundering in time-evolving transaction networks
CN109829721B (zh) 基于异质网络表征学习的线上交易多主体行为建模方法
Ofori-Boateng et al. Topological anomaly detection in dynamic multilayer blockchain networks
CN111833182B (zh) 识别风险对象的方法和装置
JP2019512128A (ja) データの秘匿性−実用性間のトレードオフを算出するためのシステムおよび方法
US20130332374A1 (en) Fraud prevention for real estate transactions
WO2023029758A1 (zh) 一种企业经济犯罪侦查方法、系统及设备
Rutskiy et al. Prospects for the Use of Artificial Intelligence to Combat Fraud in Bank Payments
Trozze et al. Of degens and defrauders: Using open-source investigative tools to investigate decentralized finance frauds and money laundering
CN112991079B (zh) 多卡共现就医欺诈行为检测方法、系统、云端及介质
Chen et al. A novel approach to uncover health care frauds through spectral analysis
Huang et al. Miracle or mirage? a measurement study of nft rug pulls
Krysovatyy et al. Classification Method of Fictitious Enterprises Based on Gaussian Naive Bayes
Dhurandhar et al. Robust system for identifying procurement fraud
Wu et al. Towards Understanding Asset Flows in Crypto Money Laundering Through the Lenses of Ethereum Heists
Onu et al. Detection of Ponzi scheme on Ethereum using machine learning algorithms
Redmond et al. Mining dense structures to uncover anomalous behaviour in financial network data

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22862899

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE