Risk analysis method and device of financial system, electronic equipment and storage medium

Abstract

The invention discloses a risk analysis method and device for a financial system, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence or other related fields, and the method comprises the steps: obtaining multi-source financial data which at least comprises structured data and unstructured text data; performing semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information; extracting risk entities, association relationships among the risk entities, risk events and attribute information of the risk events from the analysis result; constructing a financial risk knowledge graph based on the risk entities, the association relationships, the risk events and the attribute information; and performing association analysis based on the financial risk knowledge graph to obtain a risk analysis result, and executing risk early warning based on the risk analysis result. Through the method and the device, the technical problem of low risk analysis accuracy of the financial transaction system in related technologies is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence or other related fields, in particular, to a risk analysis method and device for a financial system, an electronic device and a storage medium. BACKGROUND

[0002] In the contemporary financial system, risk monitoring technology is undergoing a transformation from traditional computing methods to intelligent analysis systems. For a long time, risk management strategies based on rule engines and statistical models have been effective in dealing with historical financial data, but when faced with massive unstructured information such as news texts, social media updates, and corporate user announcements, these methods gradually reveal their limitations: the rich financial dynamics and implied risks contained in unstructured data are difficult to capture by traditional linear or static models, resulting in an increase in blind spots in the risk identification process, which in turn affects the overall risk analysis accuracy of the financial transaction system.

[0003] Moreover, existing technologies can only identify known risk patterns, and lack sufficient sensitivity to unknown or emerging risk types. Even if some risks can be identified, it is difficult to conduct in-depth causal analysis and risk propagation path tracing, making the risk warning information not specific and operable enough, and reducing the effectiveness of the warning.

[0004] In view of the above problems, no effective solutions have been proposed so far. SUMMARY

[0005] The main purpose of the present application is to provide a risk analysis method and device for a financial system, an electronic device and a storage medium, to at least solve the technical problem of low risk analysis accuracy of the financial transaction system in the related art.

[0006] In order to achieve the above purpose, according to one aspect of the present application, a risk analysis method for a financial system is provided, which comprises: acquiring multi-source financial data, wherein the multi-source financial data at least includes: structured data and unstructured text data; performing semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information; extracting risk entities, association relationships between the risk entities, risk events and attribute information of the risk events from the analysis result; constructing a financial risk knowledge graph based on the risk entities, the association relationships, the risk events and the attribute information; performing association analysis based on the financial risk knowledge graph to obtain a risk analysis result, and executing risk warning based on the risk analysis result.

[0007] Further, the step of performing semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information comprises: performing word segmentation and part-of-speech tagging on unstructured text data in the multi-source financial data to obtain text units with grammatical tags; performing field mapping and format conversion processing on structured data in the multi-source financial data to obtain standardized data fields; and performing semantic tagging on the text units and the data fields based on a financial domain knowledge base to obtain an analysis result containing semantic information, wherein the semantic information at least includes financial entities and business relationships between the financial entities.

[0008] Further, the step of extracting risk entities, association relationships between the risk entities, risk events, and attribute information of the risk events from the analysis result comprises: matching and identifying risk entities from financial entities recorded in the semantic information based on a risk entity dictionary predefined in a preset financial knowledge base; identifying grammatical dependency relationships between the risk entities based on dependency syntax analysis technology and business relationships recorded in the semantic information; determining business association of the grammatical dependency relationships to obtain the association relationships between the risk entities; and performing merging processing on duplicate risk entities and co-reference resolution processing on different expressions pointing to the same risk entity in all the association relationships.

[0009] Further, the step of extracting risk entities, association relationships between the risk entities, risk events, and attribute information of the risk events from the analysis result further comprises: identifying event expressions containing risk features from the semantic information based on an event trigger word sub-base in the preset financial knowledge base, and determining the risk events based on the event expressions; determining a time attribute and an impact degree attribute of the risk events according to context relationships in which the event expressions are recorded in the unstructured text data; determining an amount attribute and a frequency attribute of the risk events according to associated fields corresponding to the event expressions in the structured data; and generating the attribute information containing the time attribute, the impact degree attribute, the amount attribute, and the frequency attribute.

[0010] Further, the step of constructing a financial risk knowledge graph based on the risk entities, the association relationships, the risk events, and the attribute information comprises: taking the risk entities as graph nodes and taking the association relationships as edges connecting the graph nodes to generate an initial graph; establishing a mapping relationship between the risk events and corresponding graph nodes in the initial graph according to a corresponding relationship between the risk events and the risk entities; adding the attribute information of the risk events to the corresponding graph nodes as node feature attributes based on the mapping relationship; and integrating and storing the graph nodes, the edges, and the node feature attributes to obtain the financial risk knowledge graph.

[0011] Further, the step of performing correlation analysis based on the financial risk knowledge graph to obtain a risk analysis result comprises: calculating a risk indicator value of each graph node based on the node feature attribute corresponding to each graph node in the financial risk knowledge graph; determining a risk propagation path in the financial risk knowledge graph according to the risk indicator value, wherein the risk propagation path comprises risk nodes and risk edges; performing risk state marking on the risk nodes and the risk edges in the risk propagation path to obtain a marking result, wherein the marking result at least comprises a risk degree marking and a risk correlation marking generated according to a preset risk judgment rule; and generating the risk analysis result according to the risk propagation path and the marking result.

[0012] Further, the step of performing risk early warning based on the risk analysis result comprises: extracting the risk degree marking and the risk correlation marking in the risk analysis result; determining a risk early warning level based on the risk degree marking, and generating a risk propagation relationship graph based on the risk correlation marking, wherein the risk propagation relationship graph comprises risk nodes and an influence range; triggering an early warning signal based on the risk early warning level, and displaying the risk propagation relationship graph and the early warning signal through a visual interface.

[0013] To achieve the above-mentioned purpose, according to another aspect of the present application, a risk analysis device of a financial system is provided, which comprises: an acquisition unit configured to acquire multi-source financial data, wherein the multi-source financial data at least comprises structured data and unstructured text data; an analysis unit configured to perform semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information; an extraction unit configured to extract risk entities, correlation relationships between the risk entities, risk events and attribute information of the risk events from the analysis result; a construction unit configured to construct a financial risk knowledge graph based on the risk entities, the correlation relationships, the risk events and the attribute information; and an analysis unit configured to perform correlation analysis based on the financial risk knowledge graph to obtain a risk analysis result, and perform risk early warning based on the risk analysis result.

[0014] Further, the analysis unit comprises: a first processing module configured to perform word segmentation processing and part-of-speech tagging on the unstructured text data in the multi-source financial data to obtain text units with syntax markers; a second processing module configured to perform field mapping and format conversion processing on the structured data in the multi-source financial data to obtain standardized data fields; and a tagging module configured to perform semantic tagging on the text units and the data fields based on a financial domain knowledge base to obtain an analysis result containing semantic information, wherein the semantic information at least comprises financial entities and business relationships between the financial entities.

[0015] Further, the extraction unit comprises: a matching module, configured to match and identify risk entities from financial entities recorded in the semantic information based on a predefined risk entity dictionary in a preset financial knowledge base; an identification module, configured to identify a grammatical dependency relationship between the risk entities based on a dependency syntax analysis technique and a business relationship recorded in the semantic information; a judgment module, configured to perform business relevance judgment on the grammatical dependency relationship to obtain the association relationship between the risk entities; a third processing module, configured to perform merging processing on repeated risk entities, and perform co-reference resolution processing on different expressions pointing to the same risk entity in all the association relationships.

[0016] Further, the extraction unit further comprises: a first determination module, configured to identify an event expression containing a risk feature from the semantic information based on an event trigger word sub-base in the preset financial knowledge base, and determine the risk event based on the event expression; a second determination module, configured to determine a time attribute and an impact degree attribute of the risk event according to a context relationship recorded in the unstructured text data; a third determination module, configured to determine an amount attribute and a frequency attribute of the risk event according to an associated field corresponding to the event expression in the structured data; and a first generation module, configured to generate the attribute information containing the time attribute, the impact degree attribute, the amount attribute and the frequency attribute.

[0017] Further, the construction unit comprises: a second generation module, configured to generate an initial graph by taking the risk entities as graph nodes and taking the association relationships as edges connecting the graph nodes; an establishment module, configured to establish a mapping relationship between the risk event and a corresponding graph node in the initial graph according to a corresponding relationship between the risk event and the risk entity; an addition module, configured to add the attribute information of the risk event to the corresponding graph node as a node feature attribute based on the mapping relationship; and a storage module, configured to integrate and store the graph nodes, the edges and the node feature attributes to obtain the financial risk knowledge graph.

[0018] Further, the analysis unit comprises: a calculation module configured to calculate a risk indicator value of each graph node in the financial risk knowledge graph based on the node feature attribute corresponding to each graph node; a fourth determination module configured to determine a risk propagation path in the financial risk knowledge graph according to the risk indicator value, wherein the risk propagation path comprises a risk node and a risk edge; a marking module configured to mark the risk node and the risk edge in the risk propagation path to obtain a marking result, wherein the marking result at least comprises a risk degree marking and a risk association marking generated according to a preset risk judgment rule; and a third generation module configured to generate the risk analysis result according to the risk propagation path and the marking result.

[0019] Further, the analysis unit further comprises: an extraction module configured to extract the risk degree marking and the risk association marking in the risk analysis result; a fourth generation module configured to determine a risk warning level based on the risk degree marking and generate a risk propagation relationship graph based on the risk association marking, wherein the risk propagation relationship graph comprises a risk node and an influence range; and a triggering module configured to trigger a warning signal based on the risk warning level and display the risk propagation relationship graph and the warning signal through a visual interface.

[0020] To achieve the above object, according to another aspect of the present application, a computer readable storage medium is provided, which comprises a stored computer program, wherein the computer readable storage medium controls a device where the computer readable storage medium is located to execute the risk analysis method of the financial system according to any one of the above aspects when the computer program is executed.

[0021] To achieve the above object, according to another aspect of the present application, an electronic device is provided, which comprises one or more processors and a memory, wherein the memory is configured to store one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the risk analysis method of the financial system according to any one of the above aspects.

[0022] To achieve the above object, according to another aspect of the present application, a computer program product is provided, which comprises computer instructions, wherein the computer instructions are executed by a processor to implement the steps of the risk analysis method of the financial system according to any one of the above aspects.

[0023] The application provides a risk analysis method of a financial system, which comprises the following steps: acquiring multi-source financial data, wherein the multi-source financial data at least comprises structured data and unstructured text data; performing semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information; extracting risk entities, correlation between the risk entities, risk events and attribute information of the risk events from the analysis result; constructing a financial risk knowledge graph based on the risk entities, the correlation, the risk events and the attribute information; performing correlation analysis based on the financial risk knowledge graph to obtain a risk analysis result; and performing risk early warning based on the risk analysis result.

[0024] In the application, the natural language processing and graph construction technology are integrated, the multi-source financial data is deeply analyzed and the core semantics thereof is extracted, the risk entities and the complex relationships thereof are accurately identified, the technical effects of comprehensively understanding the financial market dynamics and realizing real-time early warning are achieved, and the technical problem of low risk analysis accuracy of the financial transaction system in the related art is solved.

[0025] Specifically, the application first acquires multi-source financial information comprising structured data and unstructured text data, performs deep semantic analysis on the multi-source data by using an NLP algorithm, ensures that the financial entities, events and attributes in each data segment are accurately captured, then extracts key risk entities and multi-dimensional correlation therebetween through fine processing of the analysis result, and forms a basic framework of a financial risk knowledge graph together with risk events and detailed attributes. The knowledge graph not only comprehensively covers the global perspective of the financial field, but also can reveal complex risk patterns hidden behind the data through correlation analysis of the graph. Based on the financial risk knowledge graph, correlation analysis is further performed, the graph query and the graph neural network technology are used to evaluate the risk from multiple angles, a detailed risk analysis report is formed, the accuracy and timeliness of risk early warning are greatly improved, the financial institutions can deeply understand the risk sources and formulate more accurate risk prevention and control strategies, the short board of the traditional risk analysis method in processing unstructured data is overcome, a new path is opened for intelligent risk management of the financial transaction system, and a qualitative leap in risk analysis accuracy is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0026] The accompanying drawings, which form a part of the present application, are intended to provide further understanding of the present application, and the illustrative embodiments of the present application and their description serve the purpose of explaining the present application. The accompanying drawings should not be construed in a limiting manner as to the present application.

[0027] Figure 1 Fig. 1 shows a hardware structure block diagram of a computer terminal (or a mobile device) for implementing the risk analysis method of the financial system;

[0028] Figure 2is a flow chart of a risk analysis method of an optional financial system according to an embodiment of the present application;

[0029] Figure 3 is a schematic diagram of a risk analysis device of an optional financial system according to an embodiment of the present application;

[0030] Figure 4 is a structural block diagram of an electronic device for performing a risk analysis method of a financial system according to an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to make the personnel in the technical field better understand the present application scheme, the technical scheme in the embodiment of the present application will be described clearly and completely below in combination with the drawings in the embodiment of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the scope of protection of the present application.

[0032] It should be noted that the terms "first", "second" and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0033] In order to facilitate those skilled in the art to understand the present application, the following explains some terms or nouns involved in each embodiment of the present application:

[0034] NLP, Natural Language Processing, is an artificial intelligence technology for computer to understand and generate human natural language. In the present application, NLP is used to parse unstructured text data, extract entity, relationship and event information.

[0035] Knowledge Graph, knowledge graph, is a structured database for storing and expressing the relationship between entities, events and attributes, and can perform complex queries and reasoning through graphical display of information, which is the key data structure for realizing risk correlation analysis in the present application.

[0036] GNN, Graph Neural Network, a deep learning model for processing graph structure data, in the present application, GNN is used for risk pattern recognition and association reasoning on knowledge graph, to capture complex risk patterns.

[0037] It should be noted that the risk analysis method and device of the financial system in the present application can be used in the field of artificial intelligence technology in the case of deep semantic analysis and intelligent association analysis of unstructured data of the financial market, and can also be used in any field other than the field of artificial intelligence technology in the case of deep semantic analysis and intelligent association analysis of unstructured data of the financial market. The application field of the risk analysis method and device of the financial system in the present application is not limited.

[0038] It should be noted that the related information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, storage, processing, transmission, provision, disclosure, use and processing of related data comply with the relevant laws, regulations and standards of the region, necessary security measures are taken, and the public order is not violated, and appropriate operation entrances are provided for users to choose authorization or refusal. For example, the system and the interface between the related users or institutions are provided with an interface, and before obtaining the related information, the acquisition request needs to be sent to the aforementioned user or institution through the interface, and after receiving the consent information feedback from the aforementioned user or institution, the related information is obtained.

[0039] The information collection (such as user voice, video, text collection) and analysis operation involved in the present application have provided corresponding operation entrances for the user when executing, for the user to choose to agree or refuse the automatic decision result; if the user chooses to refuse, the expert decision process is entered.

[0040] The following embodiments of the present application can be applied to various systems / applications / devices that need to perform financial risk identification and market dynamic monitoring, and can realize an intelligent risk analysis solution based on natural language processing and knowledge graph technology. The present application uses natural language processing technology to perform deep semantic analysis on unstructured text data, extracts risk-related entity, event and attribute information, and then constructs and utilizes a financial risk knowledge graph for association analysis, which can better capture complex risk signals in the market and potential influence paths between entities, while suspected risks are verified and evaluated in real time through dynamic graph updating.

[0041] The application also optimizes the risk early warning mechanism intelligently by real-time data acquisition and rapid knowledge graph updating, accurately locates potential risks, and automatically and quickly responds in the early warning process, thereby ensuring the timeliness and accuracy of risk identification and effectively supporting the risk management decision of the financial institution.

[0042] The application will be described in detail below in combination with various embodiments.

[0043] Embodiment one

[0044] According to the embodiments of the application, an embodiment of a risk analysis method of a financial system is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0045] The embodiment of the risk analysis method of the financial system provided by the embodiment of the application can be executed in a mobile terminal, a computer terminal, or a similar computing device. Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing the risk analysis method of the financial system is shown. As shown in Figure 1 , the computer terminal 10 (or mobile device) can include one or more processors 102 (the processor 102 can include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it can also include a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. Those skilled in the art can understand that Figure 1 The structure shown is only schematic, and it does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 10 can also include more or fewer components than those shown in Figure 1 , or have a different configuration from that shown in Figure 1 .

[0046] It should be noted that the one or more processors 102 and / or other data processing circuitry described above can be referred to herein generally as "data processing circuitry." The data processing circuitry can be embodied in whole or in part as software, hardware, firmware, or any combination thereof. In addition, the data processing circuitry can be a single standalone processing module, or incorporated in whole or in part within any one of the other elements of the computer terminal 10 (or mobile device). As referred to in the embodiments of the present application, the data processing circuitry functions as a processor to control, for example, the selection of the variable resistance terminal path connected to the interface.

[0047] The memory 104 can be used to store software programs of application software and modules, such as the program instructions / data storage means corresponding to the risk analysis method of the financial system in the embodiments of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, i.e. implements the risk analysis method of the financial system described above. The memory 104 can include a high-speed random access memory, and can further include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 can further include a memory disposed remotely with respect to the processor 102, which can be connected to the computer terminal 10 through a network. Examples of the network include, but are not limited to, the Internet, an enterprise user intranet, a local area network, a mobile communication network, and combinations thereof.

[0048] The transmission device 106 is used to receive or send data via a network. Specific examples of the network can include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet in a wireless manner.

[0049] The display can be, for example, a touch screen type liquid crystal display (LCD), which can enable a user to interact with the user interface of the computer terminal 10 (or mobile device).

[0050] In the above operating environment, the present application provides a method for analyzing risk of a financial system, comprising the steps of: Figure 2The financial system risk analysis method shown, the implementation subject of the method is an intelligent financial risk monitoring system, combines natural language processing (NLP) and knowledge graph construction technology, is used for financial market risk intelligent analysis and early warning scene, especially for the identification and evaluation problem of the risk implied in unstructured data, through deep semantic analysis and graph correlation analysis means, to realize the purpose of improving the accuracy and response speed of financial system risk analysis, and help financial institutions make faster and more accurate risk management decisions in complex and changing market environment.

[0051] The embodiments of the present application will be described in detail below in combination with various specific steps.

[0052] Figure 2 It is an optional flowchart of a financial system risk analysis method according to an embodiment of the present application, as shown, the method comprises the following steps: Figure 2

[0053] Step S201, acquiring multi-source financial data, wherein the multi-source financial data at least includes: structured data and unstructured text data.

[0054] Specifically, multi-source financial data refers to a collection of financial related information from different channels and forms, covering a wide range of financial activities and market dynamics. It should be noted that the diversity of multi-source financial data can provide multi-level information perspective, ensuring that the input information of the risk monitoring system is extensive and in-depth, which is beneficial to discover the risk signals hidden under the surface or spread across fields, and helps to fully understand the financial market situation and effectively monitor potential risks.

[0055] Structured data refers to data with a clear format and organization, usually stored in table form, such as numbers, dates, and information in predefined fields. In the financial field, structured data can include but is not limited to:

[0056] Transaction records: detailed transaction time, price, quantity and transaction party information;

[0057] Credit rating data: credit scores, historical default records, etc. of institutional or individual users;

[0058] Regulatory announcements: officially released rule changes, compliance requirement updates, etc.;

[0059] Economic indicators: macroeconomic data such as GDP growth rate, unemployment rate, inflation rate, etc.;

[0060] Corporate financial statements: key financial indicators such as revenue, profit, liabilities, assets, etc.

[0061] ​Unstructured text data refers to data without a predefined data model or format, represented in free-text form, covering a large number of non-numeric qualitative descriptions and analyses, including market sentiment, public opinion direction, and unexpected events, which cannot be directly quantified. Unstructured text data can come from:

[0062] News reports: market trends, corporate user news, economic rule interpretation, etc.

[0063] Social media: investor discussions, market sentiment reflections, real-time news hotspots, etc.

[0064] Corporate user reports: annual reports, social responsibility reports, management discussion and analysis, etc.

[0065] Legal documents: litigation announcements, contract terms, regulatory reports, etc.

[0066] Research analysis: industry research reports, market analysis, economic forecasts, etc.

[0067] Combining the above two types of data, the embodiments of the present application aim to build a comprehensive risk monitoring framework that can not only take advantage of the accuracy and quantification of structured data, but also extract semantic information and potential patterns from unstructured text data, thereby achieving a deeper understanding and a wider range of applications in risk identification and monitoring. Through the above data integration strategy, not only the accuracy of risk analysis is improved, but also the sensitivity to emerging or abnormal risk situations is enhanced, providing more comprehensive and timely risk warning services for financial institutions.

[0068] Step S202, performing semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information.

[0069] Specifically, semantic analysis is the process of converting natural language text into structured data, and the core purpose is to extract meaningful information segments from multi-source financial data. In the context of financial risk monitoring, semantic analysis helps to understand the implicit risk signals in unstructured text, including but not limited to market trends, corporate user activities, regulatory changes, and other key information, providing semantic-rich data support for building a knowledge graph.

[0070] In some optional embodiments, the techniques for performing semantic analysis mainly include various sub-tasks of natural language processing (NLP), and the specific steps include data preprocessing, entity recognition and relationship extraction, event extraction, and attribute extraction.

[0071] Specifically, in order to improve the efficiency and accuracy of analysis, the collected multi-source financial data needs to be preprocessed first, including data cleaning, removing irrelevant information or noise, and word segmentation and part-of-speech tagging, which divides the text into basic semantic units and tags them with grammatical labels.

[0072] Entity recognition and relation extraction are two key steps in semantic parsing, where entity recognition is used to locate entities in the text, such as corporate users, individual users, financial products, time points, etc., and relation extraction further finds the connections between these entities, such as financial relationships such as transactions, loans, ownership, etc., and key attribute information such as event time, location, etc.

[0073] In addition, in the financial field, specific events (such as mergers and acquisitions, listing, default events, etc.) often carry a lot of market impact and risk signals, and event extraction technology can identify these specific events from the text and extract their relevant details, providing important clues for risk assessment.

[0074] In addition to entities and relationships, entity attribute information in the text (such as the financial status of corporate users, the credit rating of individual users, the risk level of financial products, etc.) is also an important part of building a knowledge graph, and attribute extraction technology is used to extract these entity attribute information to refine each node in the knowledge graph.

[0075] The parsed results after semantic parsing are a series of structured data containing rich semantic information, including not only the core entities in the text, the relationships between entities, and event information, but also attribute data directly related to risk identification.

[0076] An optional specific form of the parsed results can be an entity list, which lists all the entities identified in the text, such as "Entity A is a listed corporate user", "Entity B is an investment financial institution";

[0077] The parsed results can also be a relationship set, which describes the association between entities, such as "Entity A has a loan relationship with Entity B", "Entity C conducted a merger and acquisition activity at time T";

[0078] The parsed results can also be event descriptions, which record the details of the financial events that occurred, including participants, time, location, and event type, such as "Entity D defaulted in event E";

[0079] The parsed results can also be attribute information, which lists the attributes of each entity, such as "Entity A's latest financial report shows a loss", "Entity B's credit rating is BBB+";

[0080] The above analysis results are important basis for subsequent construction of a knowledge graph and risk analysis, and the embodiment of the application converts unstructured text into structured data, which not only improves the processability of information, but also provides more dimensional data support for intelligent risk monitoring. In the subsequent steps, the analysis results containing semantic information will be integrated into the knowledge graph to construct a complex relationship network between entities, and provide a more accurate and comprehensive perspective for risk identification and evaluation.

[0081] Optionally, in the risk analysis method of the financial system provided by the embodiment of the application, the step of performing semantic analysis on the multi-source financial data to obtain the analysis results containing semantic information comprises: performing word segmentation processing and part-of-speech tagging on the unstructured text data in the multi-source financial data to obtain text units with syntax labels; performing field mapping and format conversion processing on the structured data in the multi-source financial data to obtain standardized data fields; and performing semantic tagging on the text units and the data fields based on a financial domain knowledge base to obtain the analysis results containing semantic information, wherein the semantic information at least includes financial entities and business relationships between the financial entities.

[0082] In a specific implementation scenario, when performing semantic analysis on unstructured text data, first, word segmentation processing is performed to divide the text into meaningful word units, and then part-of-speech tagging is performed to assign a syntax label, such as noun, verb, adjective, etc., to each word unit. Through word segmentation and part-of-speech tagging, the financial system can more accurately identify financial entities, business activities, and market dynamics in the text.

[0083] For structured data, such as transaction records, financial statements, credit ratings, etc., field mapping and format conversion processing are performed. Field mapping is used to ensure that data from different sources can be unified into a predefined field structure, facilitating comparison and analysis; format conversion converts data into a standard format that can be processed by the financial system, such as converting dates from various representations to the YYYY-MM-DD format, converting numerical values from text form to numerical type, etc. The above processing steps make the integration of structured data and unstructured text data more smooth, and also provide standardized data input for constructing a knowledge graph.

[0084] Another need to be explained, the processed text units and data fields are tagged with semantic information based on the professional knowledge base in the financial field, that is, each word unit or data field is assigned a deeper financial meaning, such as "loan" relationship, "bankruptcy" event, "credit rating downgrade", etc. By comparing the domain knowledge base, the financial entities (such as financial institutions, enterprise users, individual users, investors) and their business relationships (such as borrowing, guarantee, equity holding) implied in the text can be identified.

[0085] The above semantic annotation based on domain knowledge not only improves the accuracy of entity recognition and relation extraction, but also mines deeper risk signals and market trends under the surface information of the text, providing a highly structured and semantically rich dataset for subsequent risk analysis.

[0086] In step S203, risk entities, associated relationships between risk entities, risk events, and attribute information of the risk events are extracted from the analysis results.

[0087] Specifically, risk entities refer to various objects that may cause or suffer from risks in financial market activities, including but not limited to financial institutions, enterprise users, individual users, financial products, or market indexes, etc. It should be noted that these risk entities are key nodes in the risk transmission chain, and through in-depth understanding of the behavior, state, and interaction of these risk nodes, potential risk factors and market dynamics can be understood.

[0088] The associated relationships between risk entities describe how different entities influence each other, which can be direct influence such as loan relationship, equity holding, contract binding, etc., or indirect influence such as through joint participation in a market activity or being affected by the same compliance condition. Identifying the associated relationships helps to establish causal chains and risk transmission paths between entities, enabling the financial system not only to observe the risks of individual entities, but also to understand how risks spread and converge within the entire financial system, which is conducive to accurately predicting risk trends and assessing overall risk levels.

[0089] Risk events are core concepts in the field of financial risk monitoring, referring to specific events that have a significant negative impact on the market or entities, such as enterprise user bankruptcy, market volatility, significant compliance condition changes, etc. The attribute information of the event is more detailed, including the time, location, involved entities, event type, and severity of the event, etc.

[0090] In an optional embodiment, the process of extracting key risk information from the analysis results generated in step S202 in step S203 is a fine data screening and sorting step. The specific implementation can include entity filtering, relationship mapping, event classification, and attribute integration.

[0091] Among them, entity filtering refers to filtering out risk-related entities from the analysis results based on specific risk identification criteria to form a risk entity list; relationship mapping refers to mapping the identified relationships between entities to a pre-set risk association category to form a risk association matrix; event classification refers to classifying the parsed events to distinguish which are risk events, while extracting relevant attribute information of each event to build an event attribute database; attribute integration refers to integrating the detailed attribute information of risk entities and each risk entity participating in risk events to form a comprehensive risk entity file.

[0092] The embodiment of the application extracts risk entities, inter-entity correlation, risk events and attribute information, which jointly constitute a key component of the risk monitoring knowledge graph. The risk monitoring knowledge graph can not only provide instant risk early warning capability for the financial system, but also support long-term risk trend analysis and risk management decision making.

[0093] Optionally, in the risk analysis method of the financial system provided by the embodiment of the application, the step of extracting risk entities, inter-entity correlation, risk events and attribute information of the risk events from the analysis result comprises: matching and identifying risk entities from the financial entities recorded in the semantic information based on the risk entity dictionary predefined in the preset financial knowledge base; identifying the syntax dependency relationship between the risk entities based on the dependency syntax analysis technology and the business relationship recorded in the semantic information; determining the business correlation of the syntax dependency relationship to obtain the inter-entity correlation; and performing merging processing on the repeated risk entities and co-reference resolution processing on different expressions pointing to the same risk entity in all the correlation relationships.

[0094] In the processing of the analysis result containing semantic information, first, the risk entity dictionary in the preset financial knowledge base is used to match and identify whether the financial entities recorded in the text are potential risk sources. It should be noted that the risk entity dictionary contains a list of entities such as financial institutions, enterprise users, individual users and financial products that may cause or involve risks and attribute information. By comparing with the parsed text units, entities related to risk monitoring can be automatically screened out, ensuring that only entities with real risk monitoring value are focused on, and improving the analysis pertinence and efficiency.

[0095] In the step of identifying the syntax dependency relationship by the dependency syntax analysis technology, it should be noted that the dependency syntax analysis is a natural language processing technology used to analyze the dependency relationship of words in a sentence, i.e., how the words are connected with each other through the syntax structure.

[0096] In the embodiment of the application, the dependency syntax analysis technology is used to identify the syntax dependency relationship between the risk entities, such as who is the subject of the loan and who is the object of the guarantee. By analyzing the syntax structure, the relationship type between the entities can be more accurately understood and extracted.

[0097] After identifying the risk entities and their syntactic dependency relations, further business relevance determination is performed, i.e., analyzing the actual relevance of these syntactic dependency relations in the financial business scenario. For example, it is determined whether the dependency relation between "investment" and "income" implies that a specific investment behavior leads to a change in income, or whether "default" and "loss" imply economic loss due to a default event. Through business relevance determination, non-business-related dependency relations can be eliminated, and the focus is on the real reflection of the financial risk entity association, obtaining the business-related risk entity association.

[0098] Another need to be explained is the execution steps of processing repeated entities and co-reference resolution. In multi-source data processing, the same risk entity may appear in different forms of expression in different texts or data segments, such as a user of an enterprise may be mentioned as "full name", "abbreviation" or "industry alias".

[0099] In order to avoid entity repetition and relationship confusion, it is necessary to merge the repeated risk entities, ensure the uniqueness of each entity, and perform co-reference resolution processing on different expressions pointing to the same risk entity in all association relations, unify the entity identification, and keep the association relationship between entities clear and accurate. This processing step ensures the coherence and readability of the graph.

[0100] The application of dependency syntax analysis technology in the embodiment of the present application enables the financial system to deeply understand the semantic relationship in the text, effectively extracts the business association between risk entities, and through business relevance determination, it can eliminate irrelevant dependencies, focus on the substantive business relationship, reduce noise, and also enhance the business relevance and practicality of risk monitoring.

[0101] In addition, co-reference resolution and repeated entity merging processing ensure the consistency and integrity of the data, avoid analysis errors caused by data redundancy or contradictions, improve the quality of the financial risk knowledge graph, and further improve the efficiency and accuracy of risk monitoring.

[0102] In summary, the above implementation steps realize the transformation from text to entity and from dependency relation to business association in the process of solving the technical problems, greatly improving the intelligence and effectiveness of the risk analysis method.

[0103] Optionally, in the risk analysis method of the financial system provided by the embodiment of the application, the step of extracting the risk entity, the correlation between the risk entities, the risk event and the attribute information of the risk event from the analysis result further comprises: identifying the event expression containing the risk feature from the semantic information based on the event trigger word sub-library in the preset financial knowledge base, and determining the risk event based on the event expression; determining the time attribute and the impact degree attribute of the risk event according to the context relationship of the event expression recorded in the unstructured text data; determining the amount attribute and the frequency attribute of the risk event according to the correlation field corresponding to the event expression in the structured data; and generating the attribute information containing the time attribute, the impact degree attribute, the amount attribute and the frequency attribute.

[0104] It should be noted that, in the field of natural language processing, the event trigger word refers to a keyword indicating the occurrence of a specific event, such as "bankruptcy", "acquisition", "decline", etc., which is often associated with risks in financial texts.

[0105] In order to extract the risk event from the analysis result, the embodiment of the application uses the event trigger word sub-library in the preset financial knowledge base as the identification basis, and searches for the event trigger word appearing in the text data. Once a matching item is found, the context is further analyzed to confirm whether the event related to the risk actually occurs. Through the identification and positioning of the event trigger word, the preliminary screening and identification of the risk event are realized.

[0106] The unstructured text data usually contains specific time information of the event occurrence and language expressions describing the impact degree of the event. Through the analysis of the context closely connected with the event expression, the embodiment of the application can extract the time attribute (such as the specific date and time range of the event occurrence) and the impact degree attribute (such as the impact size of the event on the market or entity, positive or negative effect, etc.). Determining the time attribute helps to establish the time sequence of the event, and the impact degree attribute provides an important basis for evaluating the risk level. Both of them are the key components of the attribute information of the risk event.

[0107] The structured data, especially the financial statements and transaction records, are used to provide direct evidence about the amount scale and occurrence frequency of the risk event. By matching the event expression identified in the unstructured text with the correlation field in the structured data, the amount attribute (such as the specific amount of funds involved) and the frequency attribute (such as the repetition number or periodicity of the event) of the event can be accurately obtained. The amount attribute is used to quantify the economic impact of the risk, and the frequency attribute is used to reveal the persistence and regularity of the risk.

[0108] Finally, the time attribute and the impact degree attribute extracted from the unstructured text are combined with the amount attribute and the frequency attribute obtained from the structured data to generate complete attribute information.

[0109] The embodiment of the application can quickly locate and identify potential risk events by deep application of the event trigger word sub-library, avoid misjudgment of irrelevant texts, improve the accuracy of risk event identification, and comprehensively depict the characteristics of the risk event from multiple angles by multi-dimensional depiction of the risk characteristics, i.e., integration of the time, influence degree, amount and frequency of the event.

[0110] The combination analysis of the unstructured text and the structured data in the above steps not only increases the richness of the data, but also deepens the understanding of the system on the risk event, generates attribute information for the financial institution, helps the decision maker to make risk response strategies based on accurate data and in-depth analysis, and improves the scientificity and effectiveness of the decision.

[0111] In step S204, a financial risk knowledge graph is constructed based on the risk entity, the correlation, the risk event and the attribute information.

[0112] It should be noted that the knowledge graph is a graphical structure for representing knowledge, which stores and represents the complex connection between entities and the attribute information of the entities by taking the entity as a node and the relationship as an edge. In the field of financial risk monitoring, constructing the financial risk knowledge graph is a key step of integrating the risk entity, the correlation, the risk event and the attribute information extracted from the multi-source data into a coherent and structured knowledge system. The knowledge graph can intuitively display the correlation network between entities, identify the risk propagation path, understand the market dynamics and perform deep risk analysis and prediction.

[0113] Optionally, the process of constructing the financial risk knowledge graph involves the following key steps: entity and relationship mapping, event embedding, attribute integration, and graph updating and maintenance.

[0114] The entity and relationship mapping refers to taking the extracted risk entity as a node in the graph, connecting the correlation between these entities in the form of an edge, and clearly displaying the risk transmission chain between entities by establishing an entity-relation network;

[0115] The event embedding refers to embedding the risk event and the event attribute information into the knowledge graph, taking each event as a subgraph or forming a specific edge between entities, and taking the time, location and severity information of the event as the attributes of the edge or the labels of the node to enrich the information dimension of the graph;

[0116] The attribute integration refers to integrating the attribute information of each risk entity, such as the financial status, credit rating, market performance, etc., as additional attributes of the node, which not only enhances the description of the entity, but also supports subsequent risk quantification analysis;

[0117] The updating and maintenance of the graph is designed as a continuous process to quickly consider the changes in the financial market, and the financial system ensures the timeliness and accuracy of the knowledge by updating the graph according to the newly collected data in real time or regularly.

[0118] The embodiment of the application can realize instant monitoring of the financial market dynamics, rapid positioning of the risk source, prediction of the risk propagation path and quantification of the risk influence degree by constructing the financial risk knowledge graph. The visualization feature of the knowledge graph also makes the display of the risk information more intuitive, facilitating the quick understanding and decision-making of the analysts. In addition, the dynamic updating mechanism of the graph can continuously adapt to market changes, maintaining the real-time and effectiveness of the risk monitoring.

[0119] Optionally, in the risk analysis method of the financial system provided by the embodiment of the application, the step of constructing the financial risk knowledge graph based on the risk entities, the correlation relationship, the risk events and the attribute information comprises: taking the risk entities as graph nodes and taking the correlation relationship as edges connecting the graph nodes to generate an initial graph; establishing a mapping relationship between the risk events and the corresponding graph nodes in the initial graph according to the corresponding relationship between the risk events and the risk entities; adding the attribute information of the risk events to the corresponding graph nodes as node feature attributes based on the mapping relationship; and integrating and storing the graph nodes, the edges and the node feature attributes to obtain the financial risk knowledge graph.

[0120] In an optional embodiment, the first step of constructing the financial risk knowledge graph is to take the risk entities as nodes in the graph and take the correlation relationship between them as edges connecting the nodes to generate an initial graph framework, so that the complex relationship between entities can be visualized, facilitating analysis and understanding. For example, a financial institution (entity A) issuing a loan to another enterprise user (entity B) is represented as an edge from A to B, which not only represents the business relationship of the loan, but also can attach specific loan amount and interest rate information as the attributes of the edge.

[0121] Subsequently, the mapping of the risk events to the related nodes in the initial graph is established according to the corresponding relationship between the risk events and the risk entities. Once a certain risk event is identified to involve a specific entity, the event will establish a direct connection with the graph node where the entity is located. For example, if it is identified that “enterprise user B declares bankruptcy”, the event is mapped to the node of enterprise user B, reflecting the impact of the event on the state of enterprise user B.

[0122] Based on the mapping relationship, the attribute information (such as time, impact degree, amount and frequency) of the risk event is further added to the corresponding graph node as the characteristic attribute of the node, which enriches the information content of the node, so that each node not only represents a risk entity, but also carries the related risk event and the details of the specific impact. The addition of the node characteristic attribute enables the graph to have stronger analysis capability and interpretability, and can reveal the immediate and long-term impact of the risk event on the entity state.

[0123] Finally, the graph nodes, edges and node characteristic attributes are integrated and stored to form a complete financial risk knowledge graph. In the integration and storage process, the information of the nodes and edges is de-duplicated, corrected and improved to ensure the accuracy and consistency of all information. In addition, an efficient query and update mechanism is established to facilitate real-time monitoring of the latest market dynamics, timely adjustment of the graph structure and attribute information, and maintenance of the real-time and dynamic nature of the knowledge graph.

[0124] The embodiment of the present application effectively structures the unstructured risk information by converting entities, relationships and event attributes into graph elements, forms a knowledge graph with clear logic and clear hierarchical structure, improves the information processing efficiency and analysis accuracy, enriches the information of the node characteristic attributes, not only provides immediate feedback of the risk event, but also assists decision makers in predicting long-term risk trends and market trends, enhances the support of decision making, and makes risk management and strategic planning more precise and forward-looking.

[0125] In addition, through continuous integration, storage and updating, the knowledge graph can also be self-optimized to adapt to new risk scenarios and changes in market environment, and maintain its effectiveness in complex financial environment.

[0126] Step S205, based on the financial risk knowledge graph, correlation analysis is performed to obtain a risk analysis result, and a risk warning is executed based on the risk analysis result.

[0127] Specifically, the correlation analysis refers to identifying potential risk transmission paths and risk aggregation areas by mining the deep connections between entities, relationships and events in the financial risk knowledge graph, thereby improving the accuracy and forward-looking of risk identification. In the embodiment of the present application, the correlation analysis can rely on the following methods and technologies: graph query and pattern matching, graph neural network analysis, path analysis and community detection.

[0128] Specifically, graph query and pattern matching refers to searching in the knowledge graph for specific risk patterns using a graph query language, such as identifying entities frequently associated with a specific type of risk event; graph neural network analysis refers to applying graph neural network (GNN) technology to deep learning of nodes (entities) and edges (relationships) in the graph, identifying complex risk patterns and non-linear risk relationships; path analysis refers to analyzing the speed and strength of risk transmission and the potential spread of risks by finding the shortest path or critical path between entities in the graph; community detection refers to identifying risk clusters composed of closely connected entities through community detection of the graph, and then understanding the risk concentration phenomenon in local markets or industries.

[0129] Further, the risk analysis results are specific risk assessments and insights derived from correlation analysis, which can include: risk scores, which quantify the risk level of each entity or event; risk transmission paths, which depict the specific paths of risk transmission from the source entity to the downstream entity and the risk amplification or attenuation on each path; risk warning levels, which set warning levels for potential risk events based on risk scores and transmission analysis results, such as low, medium, and high risk warnings; risk reports and recommendations, which generate detailed analysis reports describing the current status of risk entities, predicted risk trends, and management recommendations for these risks.

[0130] Further, the risk warning mechanism uses risk analysis results to trigger pre-set alarms so that financial institutions can take timely action to mitigate or avoid risk losses. The risk warning process can include threshold setting, real-time monitoring and updating, warning message generation and sending, and warning response tracking.

[0131] Threshold setting refers to setting alert thresholds for different risk types and levels, triggering an alert when the risk score of any entity or event exceeds the threshold; real-time monitoring and updating refers to real-time monitoring of changes in risk scores in the knowledge graph, and immediately starting the warning program if an abnormal rise is detected; warning message generation and sending refers to automatically generating warning messages, including risk entities, risk types, score changes, recommended response measures, etc., and sending them to relevant responsible persons in a timely manner through email, SMS, APP notifications, etc.; warning response tracking refers to recording the sending status of the warning message and the response of the recipient for subsequent analysis and improvement of the warning mechanism.

[0132] The embodiments of this invention enable proactive monitoring and proactive management of financial risks through correlation analysis and risk early warning. Risk analysis results not only help businesses understand their own risk profiles, but also provide insights into industry-wide risk dynamics, enabling the development of more proactive risk management strategies. The risk early warning mechanism ensures that once risks are identified, key decision makers are notified immediately, facilitating rapid response and minimizing risk damage.

[0133] Combining the dynamic update characteristics of knowledge graphs and the deep learning capabilities of association analysis, the embodiments of the present invention can provide financial institutions with a continuously learning and self-improving intelligent risk monitoring platform to effectively respond to complex risk challenges in the financial market.

[0134] Optionally, in the risk analysis method for the financial system provided in an embodiment of the present invention, the step of performing association analysis based on the financial risk knowledge graph to obtain risk analysis results includes: calculating the risk index value of each graph node based on the node feature attributes corresponding to each graph node in the financial risk knowledge graph; determining the risk propagation path in the financial risk knowledge graph according to the risk index value, wherein the risk propagation path includes risk nodes and risk edges; marking the risk status of the risk nodes and risk edges in the risk propagation path to obtain a marking result, wherein the marking result at least includes a risk degree mark and a risk association mark generated according to a preset risk determination rule; and generating a risk analysis result based on the risk propagation path and the marking result.

[0135] Based on the constructed financial risk knowledge graph, we conduct an in-depth analysis of the node attributes of each graph node and calculate the corresponding risk indicator value for each graph node. These risk indicators may include but are not limited to credit risk index, market volatility, and liquidity risk level. The selection and calculation of specific indicators are based on financial expertise and real-time market data. By calculating the risk indicator value of each node, we can preliminarily quantify the independent risk level of each entity.

[0136] Next, based on the calculated node risk index values, we search for risk propagation paths within the graph. This involves identifying which graph nodes and edges form the chain of risk transmission. Determining risk propagation paths requires considering factors such as the strength of the association between nodes, the level of the risk index, and the impact of the event.

[0137] For example, if entity A has a high credit risk and affects the liquidity of entity B through a loan relationship, the path from A to B is marked as a potential risk transmission path. This analytical process is mainly used to reveal patterns in how risks spread between entities.

[0138] For each risk node and risk edge in the determined risk propagation path, a risk state label is marked according to a preset risk judgment rule to generate a marking result. The risk degree label is used to reflect the risk level, and the risk association label is used to represent the possibility and strength of risk transmission between entities.

[0139] Finally, a detailed risk analysis report is generated by comprehensively considering the risk propagation path and the marking result, which not only contains the quantitative risk indicator value and the marked risk propagation path, but also provides risk traceability analysis, risk trend prediction, and suggested risk response strategies. Through the above analysis process, the financial institution can obtain a overall view of systemic risk and in-depth insight into specific entities or events.

[0140] In summary, the deepening steps of the above correlation analysis not only solve the challenge of identifying and quantifying risks in the technical problem, but also achieve the visualization of the risk propagation process, the provision of deep risk insight, and the strategy guidance based on risk analysis, greatly improving the risk management ability and market responsiveness of the financial institution.

[0141] Optionally, in the risk analysis method of the financial system provided in the embodiment of the application, the step of executing risk warning based on the risk analysis result comprises: extracting the risk degree label and the risk association label in the risk analysis result; determining a risk warning level based on the risk degree label, and generating a risk propagation relationship diagram based on the risk association label, wherein the risk propagation relationship diagram contains risk nodes and an impact range; triggering a warning signal based on the risk warning level, and displaying the risk propagation relationship diagram and the warning signal through a visual interface.

[0142] Optionally, the risk degree label and the risk association label can be extracted from the previously generated risk analysis result as the starting point of the warning process, wherein the risk degree label is used to indicate the risk degree of a risk event or entity, and the risk association label is used to reveal information about how the risk spreads among entities.

[0143] According to the extracted risk degree label, a risk warning level is automatically determined according to a preset threshold standard, and at the same time, a risk propagation relationship diagram is generated based on the risk association label to show the mutual influence and diffusion path between risk entities. In the relationship diagram, each risk entity is represented as a node, and the risk association is represented by the edges between the nodes. The thickness and color of the edges can be used to represent the size and strength of the impact range, and the risk network effect is intuitively displayed.

[0144] Once the system determines the risk warning level, the corresponding warning signal is triggered immediately. The triggering mechanism of the warning signal needs to ensure timeliness and effectiveness to minimize risk reaction delay. At the same time, the risk propagation relationship diagram and the warning signal are displayed together through a customized visual interface to the risk control team of the financial institution.

[0145] The early warning signal on the visualization interface can be displayed in the form of a pop-up window, color coding or icon prompt, ensuring the prominent presentation of key information, while the risk propagation relationship diagram is used to provide a global perspective of risks, helping analysts quickly locate the source of risks and affected areas.

[0146] The generation of the risk propagation relationship diagram in the embodiment of the present application facilitates the intuitive understanding of the flow of risks between entities by the risk control end of the financial institution, and the combination of the early warning signal and the risk propagation relationship diagram provides decision support based on data for the financial institution, and the automatic triggering and visual display reduce the need for manual intervention, improve the efficiency and reliability of the early warning process, ensure timely action, and reduce the cost of risk processing.

[0147] In summary, by performing risk early warning based on the risk analysis result, the embodiment of the present application not only solves the challenge of responding to risk changes in real time in the technical problem, but also additionally achieves the visualization of the risk propagation process, the enhancement of data-based decision support, and the significant improvement of early warning efficiency, effectively promoting the intelligent upgrading of the risk prevention and control system of the financial institution.

[0148] Through the above steps S201 to S205, the multi-source financial data can be obtained first, wherein the multi-source financial data at least includes: structured data and unstructured text data, then the multi-source financial data is subjected to semantic analysis to obtain an analysis result containing semantic information, then the risk entities, the association relationship between the risk entities, the risk events and the attribute information of the risk events are extracted from the analysis result, then the financial risk knowledge graph is constructed based on the risk entities, the association relationship, the risk events and the attribute information, and finally the risk analysis result is obtained based on the correlation analysis of the financial risk knowledge graph, and the risk early warning is performed based on the risk analysis result.

[0149] In the embodiment of the present application, the integrated natural language processing and graph construction technology is adopted, the multi-source financial data is deeply analyzed and its core semantics is refined, the purpose of accurately identifying risk entities and their complex relationships is achieved, thereby realizing the technical effects of comprehensive insight into the financial market dynamics and real-time early warning, and further solving the technical problem of low risk analysis accuracy of the financial transaction system in the related art.

[0150] Specifically, the embodiment of the present application first acquires multi-source financial information including structured data and unstructured text data, uses an NLP algorithm to perform deep semantic analysis on these multi-source data, ensures that financial entities, events and attributes in each data segment are accurately captured, and then extracts key risk entities and their multi-dimensional association relationships through fine processing of the analysis results, together with risk events and detailed attributes to form the basic framework of the financial risk knowledge graph. This knowledge graph not only includes the overall perspective of the financial field, but also can reveal complex risk patterns hidden behind the data through association analysis of the graph; based on the financial risk knowledge graph, further association analysis is performed, and through graph query and graph neural network technology, the risk is evaluated from multiple angles to form a detailed risk analysis report, greatly improving the accuracy and timeliness of risk early warning, helping financial institutions to deeply understand the risk sources and develop more accurate risk prevention strategies, overcoming the short board of traditional risk analysis methods in processing unstructured data, opening up a new path for intelligent risk management of financial transaction systems, and realizing a qualitative leap in risk analysis accuracy.

[0151] The present application will be described below in conjunction with another alternative embodiment.

[0152] Embodiment two

[0153] The embodiment of the present application also provides a risk analysis device for a financial system. It should be noted that the risk analysis device for the financial system of the embodiment of the present application comprises a plurality of implementation units and can be used to execute the risk analysis method for the financial system provided in the above embodiment one. Each implementation unit corresponds to each implementation step in the above embodiment one.

[0154] Figure 3 is a schematic diagram of an alternative risk analysis device for a financial system according to the embodiment of the present application, as shown in the figure, the device can comprise: an acquisition unit 31, an analysis unit 32, an extraction unit 33, a construction unit 34 and an analysis unit 35. Figure 3

[0155] The acquisition unit 31 is configured to acquire multi-source financial data, wherein the multi-source financial data at least comprises structured data and unstructured text data.

[0156] The analysis unit 32 is configured to perform semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information.

[0157] The extraction unit 33 is configured to extract risk entities, association relationships between risk entities, risk events and attribute information of the risk events from the analysis result.

[0158] ​The constructing unit 34 is configured to construct the financial risk knowledge graph based on the risk entity, the correlation, the risk event and the attribute information.

[0159] The analyzing unit 35 is configured to perform correlation analysis based on the financial risk knowledge graph to obtain a risk analysis result, and perform risk early warning based on the risk analysis result.

[0160] The risk analysis device of the financial system can first acquire multi-source financial data through the acquiring unit 31, wherein the multi-source financial data at least includes structured data and unstructured text data, then perform semantic analysis on the multi-source financial data through the analyzing unit 32 to obtain an analysis result containing semantic information, then extract the risk entity, the correlation between the risk entities, the risk event and the attribute information of the risk event from the analysis result through the extracting unit 33, then construct the financial risk knowledge graph based on the risk entity, the correlation, the risk event and the attribute information through the constructing unit 34, and finally perform correlation analysis based on the financial risk knowledge graph through the analyzing unit 35 to obtain a risk analysis result, and perform risk early warning based on the risk analysis result.

[0161] In the embodiment of the present application, the integrated natural language processing and graph construction technology is adopted, the core semantics of the multi-source financial data is extracted through deep analysis, the risk entity and its complex relationship are accurately identified, the comprehensive insight into the financial market dynamics and the real-time early warning are realized, and the technical problem of low risk analysis accuracy of the financial transaction system in the related art is solved.

[0162] Specifically, the multi-source financial information including structured data and unstructured text data is first acquired, the multi-source data is deeply analyzed through the NLP algorithm, the financial entity, event and attribute in each data segment are accurately captured, then the key risk entity and the multi-dimensional correlation between them are extracted through fine processing of the analysis result, and the risk event and detailed attribute together constitute the basic framework of the financial risk knowledge graph, the knowledge graph not only includes the global perspective of the financial field, but also can reveal the complex risk pattern hidden behind the data through correlation analysis of the graph, the correlation analysis is further implemented based on the financial risk knowledge graph, the risk is evaluated from multiple angles through graph query and graph neural network technology, a detailed risk analysis report is formed, the accuracy and timeliness of the risk early warning are greatly improved, the financial institutions can deeply understand the risk source and develop more accurate risk prevention and control strategies, the short board of the traditional risk analysis method in processing unstructured data is overcome, a new path for intelligent risk management of the financial transaction system is opened, and a qualitative leap in risk analysis accuracy is realized.

[0163] Further, the parsing unit comprises: a first processing module configured to perform word segmentation processing and part-of-speech tagging on the unstructured text data in the multi-source financial data to obtain text units with syntax labels; a second processing module configured to perform field mapping and format conversion processing on the structured data in the multi-source financial data to obtain standardized data fields; and a labeling module configured to perform semantic labeling on the text units and the data fields based on a financial domain knowledge base to obtain a parsing result containing semantic information, wherein the semantic information at least includes financial entities and business relationships between the financial entities.

[0164] Further, the extraction unit comprises: a matching module configured to match and identify risk entities from the financial entities recorded in the semantic information based on a predefined risk entity dictionary in the preset financial knowledge base; an identification module configured to identify syntax dependency relationships between the risk entities based on dependency syntax analysis technology and the business relationships recorded in the semantic information; a determination module configured to perform business correlation determination on the syntax dependency relationships to obtain correlation relationships between the risk entities; and a third processing module configured to perform merging processing on repeated risk entities and co-reference resolution processing on different expressions pointing to the same risk entity in all the correlation relationships.

[0165] Further, the extraction unit further comprises: a first determination module configured to identify event expressions containing risk features from the semantic information based on an event trigger sub-base in the preset financial knowledge base, and determine risk events based on the event expressions; a second determination module configured to determine time attribute and impact degree attribute of the risk events according to context relationships of the event expressions recorded in the unstructured text data; and a third determination module configured to determine amount attribute and frequency attribute of the risk events according to associated fields corresponding to the event expressions in the structured data; and a first generation module configured to generate attribute information containing the time attribute, the impact degree attribute, the amount attribute and the frequency attribute.

[0166] Further, the construction unit comprises: a second generation module configured to generate an initial graph by taking the risk entities as graph nodes and taking the correlation relationships as edges connecting the graph nodes; an establishment module configured to establish a mapping relationship between the risk events and corresponding graph nodes in the initial graph according to a corresponding relationship between the risk events and the risk entities; an addition module configured to add the attribute information of the risk events to the corresponding graph nodes as node feature attributes based on the mapping relationship; and a storage module configured to integrate and store the graph nodes, the edges and the node feature attributes to obtain the financial risk knowledge graph.

[0167] Further, the analysis unit comprises: a calculation module configured to calculate a risk index value of each graph node in the financial risk knowledge graph based on the node feature attribute corresponding to each graph node; a fourth determination module configured to determine a risk propagation path in the financial risk knowledge graph according to the risk index value, wherein the risk propagation path comprises a risk node and a risk edge; a marking module configured to mark the risk node and the risk edge in the risk propagation path to obtain a marking result, wherein the marking result at least comprises a risk degree marking and a risk association marking generated according to a preset risk judgment rule; and a third generation module configured to generate a risk analysis result according to the risk propagation path and the marking result.

[0168] Further, the analysis unit further comprises: an extraction module configured to extract the risk degree marking and the risk association marking in the risk analysis result; a fourth generation module configured to determine a risk warning level based on the risk degree marking and generate a risk propagation relationship graph based on the risk association marking, wherein the risk propagation relationship graph comprises a risk node and an influence range; and a triggering module configured to trigger a warning signal based on the risk warning level and display the risk propagation relationship graph and the warning signal through a visual interface.

[0169] It should be noted that the above-mentioned acquisition unit 31, analysis unit 32, extraction unit 33, construction unit 34, and analysis unit 35 correspond to steps S201 to S205 in Embodiment One, and the above-mentioned units and corresponding steps achieve the same instances and application scenarios, but are not limited to the content disclosed in Embodiment One. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in a memory (for example, memory 104) and processed by one or more processors (for example, processors 102a, 102b, …, 102n), or the above-mentioned modules or units can be a part of the device and can run in the computer terminal 10 provided in Embodiment One.

[0170] The application will be described in detail below in conjunction with another alternative embodiment.

[0171] Embodiment Three

[0172] The embodiments of the application can also provide an electronic device, Figure 4 which is a structural block diagram of an electronic device for performing a risk analysis method of a financial system according to an embodiment of the application, as Figure 4 shown, the electronic device can include one or more (only one is shown in the figure) processors 402, a memory 404, a storage controller, and a peripheral interface, wherein the peripheral interface is connected with a radio frequency module, an audio module, and a display. Figure 4

[0173] ​The memory can be configured to store software programs and modules, such as program instructions / modules corresponding to the risk analysis method and device of the financial system in the embodiments of the present application. The processor executes various functions and data processing by running the software programs and modules stored in the memory, that is, implements the above-mentioned risk analysis method of the financial system. The memory can include a high-speed random access memory, and can also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some examples, the memory can further include a memory remotely arranged with respect to the processor, which can be connected to the terminal through a network. Examples of the above-mentioned network include but are not limited to the Internet, an enterprise user intranet, a local area network, a mobile communication network, and a combination thereof.

[0174] The processor can call information and application programs stored in the memory through the transmission device to perform the following steps: acquiring multi-source financial data, wherein the multi-source financial data at least includes structured data and unstructured text data; performing semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information; extracting risk entities, association relationships between risk entities, risk events, and attribute information of the risk events from the analysis result; constructing a financial risk knowledge graph based on the risk entities, the association relationships, the risk events, and the attribute information; performing association analysis based on the financial risk knowledge graph to obtain a risk analysis result, and performing risk early warning based on the risk analysis result.

[0175] The processor can also call information and application programs stored in the memory through the transmission device to perform the following steps: performing word segmentation processing and part-of-speech tagging on the unstructured text data in the multi-source financial data to obtain text units with syntax labels; performing field mapping and format conversion processing on the structured data in the multi-source financial data to obtain standardized data fields; performing semantic tagging on the text units and the data fields based on a financial domain knowledge base to obtain an analysis result containing semantic information, wherein the semantic information at least includes financial entities and business relationships between the financial entities.

[0176] The processor can also call information and application programs stored in the memory through the transmission device to perform the following steps: matching and identifying risk entities from financial entities recorded in the semantic information based on a risk entity dictionary predefined in a preset financial knowledge base; identifying syntax dependency relationships between the risk entities based on dependency syntax analysis technology and business relationships recorded in the semantic information; performing business association determination on the syntax dependency relationships to obtain association relationships between the risk entities; performing merging processing on repeated risk entities, and performing co-reference resolution processing on different expressions pointing to the same risk entity in all association relationships.

[0177] The processor can further call information and application programs stored in the memory through the transmission device to perform the following steps: identifying an event expression containing a risk feature from the semantic information based on an event trigger sub-library in a preset financial knowledge base, and determining a risk event based on the event expression; determining a time attribute and an impact degree attribute of the risk event according to a context relationship of the event expression recorded in the unstructured text data; determining an amount attribute and a frequency attribute of the risk event according to an associated field corresponding to the event expression in the structured data; and generating attribute information containing the time attribute, the impact degree attribute, the amount attribute and the frequency attribute.

[0178] The processor can further call information and application programs stored in the memory through the transmission device to perform the following steps: generating an initial graph by taking the risk entity as a graph node and taking the association relationship as an edge connecting the graph nodes; establishing a mapping relationship between the risk event and the corresponding graph node in the initial graph according to the corresponding relationship between the risk event and the risk entity; adding the attribute information of the risk event to the corresponding graph node as a node characteristic attribute based on the mapping relationship; and integrating and storing the graph node, the edge and the node characteristic attribute to obtain a financial risk knowledge graph.

[0179] The processor can further call information and application programs stored in the memory through the transmission device to perform the following steps: calculating a risk index value of each graph node based on the node characteristic attribute corresponding to each graph node in the financial risk knowledge graph; determining a risk propagation path in the financial risk knowledge graph according to the risk index value, wherein the risk propagation path contains a risk node and a risk edge; marking the risk node and the risk edge in the risk propagation path to obtain a marking result, wherein the marking result at least contains a risk degree marking and a risk association marking generated according to a preset risk determination rule; and generating a risk analysis result according to the risk propagation path and the marking result.

[0180] The processor can further call information and application programs stored in the memory through the transmission device to perform the following steps: extracting the risk degree marking and the risk association marking in the risk analysis result; determining a risk warning level based on the risk degree marking, and generating a risk propagation relationship graph based on the risk association marking, wherein the risk propagation relationship graph contains a risk node and an impact range; triggering a warning signal based on the risk warning level, and displaying the risk propagation relationship graph and the warning signal through a visual interface.

[0181] The embodiment of the present application provides a risk analysis scheme of a financial system. By means of integrating natural language processing and graph construction technology, through deep analysis of multi-source financial data and extraction of core semantics, the purpose of accurately identifying risk entities and complex relationships is achieved, so that the technical effects of comprehensive insight into financial market dynamics and real-time early warning are realized, and the technical problem of low risk analysis accuracy of the financial transaction system in the related art is solved.

[0182] Those skilled in the art can understand that, Figure 4 The structure shown is only schematic, and the electronic device can also be a smart phone, a tablet computer, a palm computer, a mobile Internet device (MID), a PAD, and the like. Figure 4 It does not limit the structure of the electronic device. For example, the electronic device can further include more or less components (such as a network interface, a display device, and the like) than those shown in the figure, or have a different configuration from that shown in the figure. Figure 4 Figure 4 The skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by a program instructing the related hardware of the terminal device, and the program can be stored in a computer readable storage medium, which can include a flash disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and the like.

[0183] The present application will be described below in combination with another optional embodiment.

[0184] The present application will be described below in combination with another optional embodiment.

[0185] Embodiment Four

[0186] The embodiment of the present application further provides a computer readable storage medium. Optionally, in the embodiment of the present application, the computer readable storage medium can be used to save the program code executed by the risk analysis method of the financial system provided in the embodiment one.

[0187] Optionally, in the embodiment of the present application, the storage medium can be located in any one of the computer terminals in the computer terminal group in the computer network, or in any one of the mobile terminals in the mobile terminal group.

[0188] ​The embodiment of the present application further provides a computer program product, when executed on a data processing device, is suitable for executing the steps of the risk analysis method of the financial system: acquiring multi-source financial data, wherein the multi-source financial data at least comprises: structured data and unstructured text data; performing semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information; extracting a risk entity, a correlation relationship between risk entities, a risk event and attribute information of the risk event from the analysis result; constructing a financial risk knowledge graph based on the risk entity, the correlation relationship, the risk event and the attribute information; performing correlation analysis based on the financial risk knowledge graph to obtain a risk analysis result, and performing risk early warning based on the risk analysis result.

[0189] The above sequence numbers of the embodiments of the present application are only for description, and do not represent advantages or disadvantages of the embodiments.

[0190] In the above embodiments of the present application, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0191] In the several embodiments provided by the present application, it should be understood that the disclosed technology can be implemented in other ways. Of course, the unit embodiment described above is only schematic. For example, the division of the units is only a logical function division. There can be another division for actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling or direct coupling or communication connection between each set can be indirect coupling or communication connection through some interface, electrical or other form.

[0192] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, they can be located in one place, or distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0193] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware or in the form of software functional unit.

[0194] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal user computer, a server or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0195] The above description is only the preferred embodiments of the present application. It should be pointed out that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should be considered as the protection scope of the present application.

Claims

1. A risk analysis method for a financial system, characterized in that: include: Acquiring multi-source financial data, wherein the multi-source financial data includes at least: structured data and unstructured text data; Performing semantic analysis on the multi-source financial data to obtain an analysis result containing semantic information; Extracting risk entities, relationships between risk entities, risk events, and attribute information of risk events from the analysis results; Constructing a financial risk knowledge graph based on the risk entities, the association relationships, the risk events, and the attribute information; An association analysis is performed based on the financial risk knowledge graph to obtain a risk analysis result, and a risk warning is executed based on the risk analysis result.

2. The risk analysis method according to claim 1, characterized in that: The step of performing semantic parsing on the multi-source financial data to obtain a parsing result containing semantic information includes: Performing word segmentation and part-of-speech tagging on the unstructured text data in the multi-source financial data to obtain text units with grammatical markers; Performing field mapping and format conversion processing on the structured data in the multi-source financial data to obtain standardized data fields; The text unit and the data field are semantically annotated based on a financial domain knowledge base to obtain a parsing result containing semantic information, wherein the semantic information at least includes financial entities and business relationships between the financial entities.

3. The risk analysis method according to claim 1, characterized in that: The step of extracting risk entities, associations between the risk entities, risk events, and attribute information of the risk events from the analysis results includes: Matching and identifying risk entities from the financial entities recorded in the semantic information based on a risk entity dictionary predefined in a preset financial knowledge base; Identifying grammatical dependencies between the risk entities based on dependency parsing technology and the business relationships recorded in the semantic information; Performing business relevance determination on the grammatical dependency relationship to obtain the association relationship between the risk entities; The repeated risk entities are merged, and the different expressions pointing to the same risk entity in all the association relationships are subjected to coreference resolution.

4. The risk analysis method according to claim 1, characterized in that: The step of extracting risk entities, associations between the risk entities, risk events, and attribute information of the risk events from the analysis results further includes: identifying event descriptions containing risk characteristics from the semantic information based on an event trigger word sub-library in a preset financial knowledge base, and determining the risk event based on the event descriptions; Determining the time attribute and impact degree attribute of the risk event based on the contextual relationship of the event description recorded in the unstructured text data; Determining the amount attribute and frequency attribute of the risk event according to the associated fields corresponding to the event description in the structured data; The attribute information including the time attribute, the influence degree attribute, the amount attribute, and the frequency attribute is generated.

5. The risk analysis method according to claim 1, characterized in that: The step of constructing a financial risk knowledge graph based on the risk entities, the association relationships, the risk events, and the attribute information includes: The risk entities are used as graph nodes, and the association relationships are used as edges connecting the graph nodes to generate an initial graph; According to the corresponding relationship between the risk event and the risk entity, a mapping relationship between the risk event and the corresponding graph node in the initial graph is established; Based on the mapping relationship, the attribute information of the risk event is added to the corresponding graph node as a node feature attribute; The graph nodes, the edges, and the node feature attributes are integrated and stored to obtain the financial risk knowledge graph.

6. The risk analysis method according to claim 5, characterized in that: The steps of performing association analysis based on the financial risk knowledge graph to obtain risk analysis results include: Calculating a risk index value for each graph node based on the node characteristic attribute corresponding to each graph node in the financial risk knowledge graph; Determining a risk propagation path in the financial risk knowledge graph according to the risk indicator value, wherein the risk propagation path includes risk nodes and risk edges; Marking the risk status of the risk nodes and the risk edges in the risk propagation path to obtain a marking result, wherein the marking result at least includes a risk degree mark and a risk association mark generated according to a preset risk determination rule; The risk analysis result is generated according to the risk propagation path and the marking result.

7. The risk analysis method according to claim 1, characterized in that: The step of executing risk warning based on the risk analysis result includes: Extracting risk level markers and risk association markers from the risk analysis results; Determining a risk warning level based on the risk degree mark, and generating a risk propagation relationship diagram based on the risk association mark, wherein the risk propagation relationship diagram includes risk nodes and impact ranges; A warning signal is triggered based on the risk warning level, and the risk propagation relationship diagram and the warning signal are displayed through a visual interface.

8. A risk analysis device for a financial system, characterized in that: include: an acquisition unit, configured to acquire multi-source financial data, wherein the multi-source financial data includes at least: structured data and unstructured text data; a parsing unit, configured to perform semantic parsing on the multi-source financial data to obtain a parsing result containing semantic information; an extraction unit, configured to extract risk entities, associations between the risk entities, risk events, and attribute information of the risk events from the analysis results; A construction unit, configured to construct a financial risk knowledge graph based on the risk entities, the association relationships, the risk events, and the attribute information; An analysis unit is used to perform association analysis based on the financial risk knowledge graph to obtain risk analysis results, and to perform risk warning based on the risk analysis results.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the risk analysis method for a financial system according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the risk analysis method for the financial system as described in any one of claims 1 to 7.

11. A computer program product, characterized in that The method comprises computer instructions, wherein when the computer instructions are executed by a processor, the steps of the risk analysis method for a financial system according to any one of claims 1 to 7 are implemented.

Images