Enterprise investment and financing risk analysis method, system and device and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and relates to an enterprise investment and financing risk analysis method, system and device, and a storage medium. The enterprise investment and financing risk analysis method comprises the steps of obtaining original enterprise investment and financing data; analyzing the original enterprise investment and financing data to obtain graph demand information of enterprise investment risks; constructing an original enterprise knowledge graph according to the graph demand information, wherein the original enterprise knowledge graph comprises a plurality of nodes and edges connected with the nodes; respectively mapping the nodes and the edges to a text semantic space to obtain node semantic vectors and edge semantic vectors; updating representation of nodes in the original enterprise knowledge graph according to the node semantic vector and the edge semantic vector to obtain a new enterprise knowledge graph; and analyzing the new enterprise knowledge graph to obtain a risk analysis result of the original enterprise investment and financing data. According to the invention, the accuracy and convenience of the risk identification result of the original enterprise investment and financing data can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, system, device and storage medium for analyzing enterprise investment and financing risks. Background Art

[0002] In recent years, artificial intelligence technology has been widely used in the field of enterprise risk assessment, especially in investment and financing decisions. The enterprise risk assessment models commonly used in existing technologies include recurrent neural networks (RNN), graph neural networks (GNN) and large language models (LLM). Specifically, recurrent neural networks (RNN) extract features from corporate financial reports, news reports and social media data to predict potential risks; graph neural networks (GNN) mine potential risk information in corporate networks by modeling corporate relationship graphs, such as equity investments, business collaborations and supply chain connections; large language models (LLM) are used to process unstructured text data such as policy documents, news reports and social media posts to extract key information related to risks.

[0003] However, existing technologies still face the following key challenges in practical applications: recurrent neural networks (RNNs) have difficulty processing complex relationships between enterprises and cannot fully utilize structured data; graph neural networks (GNNs) have difficulty processing unstructured text data and lack semantic understanding capabilities; large language models (LLMs) have difficulty directly modeling topological relationships between enterprises and have limitations in processing structured data. Summary of the Invention

[0004] This application aims to at least solve the technical problems existing in the prior art and provide a method, system, device and storage medium for enterprise investment and financing risk analysis.

[0005] In a first aspect, the present invention provides a method for analyzing enterprise investment and financing risks, comprising:

[0006] Obtain original enterprise investment and financing data;

[0007] Analyze the original enterprise investment and financing data to obtain the map demand information of enterprise investment risks;

[0008] Construct an original enterprise knowledge graph based on graph demand information. The original enterprise knowledge graph includes several nodes and several edges connecting the nodes. The nodes represent keywords in the original enterprise investment and financing data, and the edges are used to represent the association relationship between the nodes.

[0009] Map nodes and edges to text semantic space respectively to obtain node semantic vectors and edge semantic vectors;

[0010] Update the representation of nodes in the original enterprise knowledge graph according to the node semantic vector and edge semantic vector to obtain a new enterprise knowledge graph;

[0011] Analyze the new enterprise knowledge graph to obtain risk analysis results of the original enterprise investment and financing data.

[0012] In a second aspect, the present invention provides an enterprise investment and financing risk analysis system, the system comprising:

[0013] Acquisition module, used to obtain original enterprise investment and financing data;

[0014] The parsing module is used to parse the original enterprise investment and financing data to obtain the graph demand information of the enterprise investment risk;

[0015] The knowledge graph construction module is used to construct the original enterprise knowledge graph based on the graph requirement information. The original enterprise knowledge graph includes several nodes and several edges connecting the nodes. The nodes represent keywords in the original enterprise investment and financing data, and the edges are used to represent the association relationship between the nodes.

[0016] A processing module is used to map nodes and edges to text semantic space to obtain node semantic vectors and edge semantic vectors respectively;

[0017] An updating module is used to update the representation of nodes in the original enterprise knowledge graph according to the node semantic vector and the edge semantic vector to obtain a new enterprise knowledge graph;

[0018] The output module is used to parse the new enterprise knowledge graph and obtain the risk analysis results of the original enterprise investment and financing data.

[0019] In a third aspect, the present invention provides an electronic device, comprising:

[0020] at least one processor; and,

[0021] a memory communicatively connected to the at least one processor; wherein,

[0022] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the above-mentioned enterprise investment and financing risk analysis method.

[0023] In a fourth aspect, the present invention also provides a computer-readable storage medium, in which at least one computer program is stored. The at least one computer program is executed by a processor in an electronic device to implement the above-mentioned enterprise investment and financing risk analysis method.

[0024] In summary, this application has the following beneficial technical effects:

[0025] The data from different fields in the original enterprise investment and financing data are parsed out and represented in the form of nodes and edges. The multi-category and multi-granularity prediction tasks are simplified into a unified "focus node" prediction, reducing the complexity of the model and enhancing knowledge sharing; the nodes and edges are mapped to the text semantic space, and the representation of the nodes in the original enterprise knowledge graph is updated according to the node semantic vector and edge semantic vector, enhancing the semantic understanding and relational reasoning ability of the graph convolutional network to obtain a new enterprise knowledge graph; the graph convolutional model can perform task prediction based on the new enterprise knowledge graph to obtain risk analysis results for the original enterprise investment and financing data, which can improve the accuracy of the model in risk identification results for multi-structure type data. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A flow chart of a method for analyzing enterprise investment and financing risks provided by one embodiment of the present invention;

[0027] Figure 2 A schematic diagram of the system structure of a method for analyzing enterprise investment and financing risks provided by one embodiment of the present invention;

[0028] Figure 3 A schematic diagram of the structure of a task-independent graph provided by one embodiment of the present invention;

[0029] Figure 4 A schematic diagram of the structure of an electronic device for implementing the enterprise investment and financing risk analysis method provided by one embodiment of the present invention.

[0030] Reference numerals: 10, processor; 11, memory; 12, communication bus; 13, communication interface.

[0031] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0032] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0033] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0034] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal communication between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0035] Reference Figure 1 FIG. 1 is a flow chart of a method for analyzing enterprise investment and financing risks according to an embodiment of the present invention. In this embodiment, the method for analyzing enterprise investment and financing risks includes:

[0036] S1. Obtain original enterprise investment and financing data.

[0037] Original enterprise investment and financing data includes structured data (such as financial data) and semi-structured data (such as Internet data).

[0038] Among them, structured financial data includes corporate financial statements, credit records and administrative penalty records. Corporate financial statements include balance sheets, income statements, cash flow statements, etc. The statements provide detailed information on the company's financial status, operating results and cash flow; credit records include the company's credit rating, loan records, repayment records, etc., reflecting the company's credit status and financial health; administrative penalty records include administrative penalty information imposed on companies for violating laws and regulations, which can help identify potential legal risks.

[0039] Semi-structured internet data includes news reports, social media comments, and legal documents. News reports, including reports from major news websites, cover a company's business dynamics, market performance, and major events. Social media comments, including user comments from social media platforms like Weibo, WeChat, and forums, can reflect the public's attitudes and opinions about a company and its products. Legal documents, including court judgments and arbitration awards, can help identify legal disputes and potential legal risks for a company.

[0040] S2. Analyze the original enterprise investment and financing data to obtain the map demand information of enterprise investment risks.

[0041] Reference Figure 2 In a preferred implementation of this embodiment, before parsing the original enterprise investment and financing data to obtain the enterprise investment risk graph demand information, the enterprise investment and financing risk analysis method further includes:

[0042] Unifying the data format of original enterprise investment and financing data, and standardizing the original enterprise investment and financing data after the unified data format to obtain standardized enterprise investment and financing data;

[0043] Based on standardized enterprise investment and financing data, we can obtain the map demand information of enterprise investment risks.

[0044] To make raw corporate investment and financing data suitable for input into large language models, it is typically necessary to cleanse, format, and standardize the data. Specifically, data cleaning removes irrelevant information, duplicate data, and erroneous data, and fills or deletes missing data to ensure data integrity.

[0045] When formatting and standardizing raw corporate investment and financing data, we first unify the data formats from different sources. For example, we standardize date and numeric formats into a standard format and unify field names, such as combining "total assets" and "total assets" into "total assets." We then standardize the formatted data to meet the input requirements of large language models. For example, we convert text data into vector representations and normalize numeric data to a specific range.

[0046] After obtaining the standardized enterprise investment and financing data, the pre-trained large-scale language model (such as BERT, GPT, etc.) is used to convert the original enterprise investment text data into semantic vectors. Specifically, the large-scale language model first extracts the language descriptions of several keywords related to enterprise investment risks (such as enterprise entities, risk events, enterprise partnerships, equity relationships, risk associations, etc.) from the standardized enterprise investment and financing data, and uses the encoder part of the large language model (LLM) to extract the text features of the language descriptions of the keywords, and maps the text features into fixed-length semantic vectors to obtain graph demand information, effectively solving the challenges of data heterogeneity in enterprise investment promotion and investment and financing risk assessment.

[0047] S3. Build the original enterprise knowledge graph based on graph demand information.

[0048] The original enterprise knowledge graph includes several nodes and several edges connecting the nodes. The nodes represent keywords in the original enterprise investment and financing data, and the edges are used to represent the association relationship between the nodes.

[0049] The original enterprise knowledge graph is a dynamic heterogeneous network graph constructed from multi-source heterogeneous data (such as corporate financial statements, judicial litigation records, industrial chain relationships, news and public opinion, etc.). Its core function is to represent the characteristics of enterprises in different dimensions and their complex interactive relationships through nodes and edges. Specifically, node types include enterprise entities (such as "Enterprise A", "Enterprise B"), related events (such as "labor disputes", "administrative penalties"), and node attributes (such as "total assets: 150 million yuan", "cooperative projects: 5"), etc. Edges include inter-enterprise relationships (such as cooperative relationships, equity relationships), event associations (such as risk associations), and edge attribute descriptions (such as "Enterprise A holds 30% of the shares of Enterprise B").

[0050] Specifically, the original enterprise knowledge graph is constructed based on the graph requirement information, including:

[0051] S31. Determine graph data information according to graph requirement information. The graph data information includes attribute names of nodes, attribute values ​​of nodes, and edge information.

[0052] In this embodiment, the attribute name refers to the node feature category (such as total assets), and the attribute value is a specific numerical value (such as 150 million yuan); the node attribute is determined by the node's attribute name and the node's attribute value. The node attribute can represent the specific characteristics of an enterprise, event or attribute, and is usually presented as structured or semi-structured data: among them, structured node attributes include enterprise nodes and risk event nodes. Enterprise nodes include total assets (numerical type), social security contribution ratio (percentage), and credit rating (type); risk event nodes include event type (administrative penalty / labor dispute), penalty amount (numerical type) and effective date (time type); semi-structured text attributes are enterprise feature descriptions such as "total assets are 150 million yuan, social security contribution ratio is 80%, and employee educational level distribution is 60% undergraduate and 25% master's degree;".

[0053] Edge information is used to describe the type and strength of relationships between nodes. Edge information includes the number of cooperative projects (numeric), duration of cooperation (time), equity share / shareholding ratio (percentage), and the cooperative relationship between enterprise A and enterprise B (such as cooperation type, number of cooperative projects, and years of cooperation).

[0054] S32. Use the graph convolutional network to construct the nodes and edges in the original enterprise knowledge graph into graph objects.

[0055] Specifically, the graph object is a task-independent graph (TAG), which only includes the node and edge structures corresponding to the graph requirement information, but does not include the specific content and edge information of the node representation. The node representation includes the node's neighbor structure and the node's semantic information. The structure of the task-independent graph can be referred to Figure 3 , Figure 3The IN tag in the task-independent graph refers to the label of the risk domain corresponding to the task-independent graph (risk domains include labor risk, debt risk, and so on). Task-independent graphs from different domains are uniformly mapped into a semantic space, and node / edge vectors are inherently compatible across tasks without the need for additional standardization. Using the Task-Independent Graph (TAG) representation enables model transfer between different domains (such as labor risk and financial risk), minimizing cross-task performance losses in task prediction models corresponding to graph objects.

[0056] S33. Determine the representation of each node according to the attribute name, attribute value and edge information of each node in the graph object to obtain the original enterprise knowledge graph.

[0057] The original enterprise knowledge graph uses task-independent graph representation to integrate different fields (such as labor risks and financial risks); the node representation is determined according to the attribute name, attribute value and edge information corresponding to the node, and the semantic vector of each node and edge is used as the representation of the node and edge of the graph. The graph structure is constructed according to the relationship between the nodes and edges to ensure the integrity and consistency of the graph structure of the original enterprise knowledge graph.

[0058] S4. Map the nodes and edges to the text semantic space respectively to obtain the node semantic vector and edge semantic vector.

[0059] Specifically, nodes and edges are mapped to the text semantic space to obtain node semantic vectors and edge semantic vectors, including:

[0060] S41. Extract the attribute name and attribute value of the node from the original enterprise investment and financing data to obtain the node attribute.

[0061] S42. Extract edge information from the original enterprise investment and financing data to obtain edge attributes.

[0062] S43. Convert the node attributes into a natural language text description to obtain a text description of the node that is readable by the language model.

[0063] S44. Convert the edge attributes into a natural language text description to obtain a text description of the edge that is readable by the language large model.

[0064] S45. Semantically encode the text description of the node and the text description of the edge respectively to obtain a node semantic vector and an edge semantic vector.

[0065] Specifically, the text description of the node and the text description of the edge are semantically encoded to obtain the node semantic vector and the edge semantic vector, including:

[0066] S451. Hierarchically extract multiple levels of node semantic features from the text description of the node, fuse the multiple levels of node semantic features, and obtain a node semantic vector.

[0067] Specifically, multiple levels of node semantic features are obtained by encoding the node's text description at the attribute level and the relationship level. When encoding the node's text description at the attribute level, Qwen2's Token Embedding layer is used to capture basic semantics. The attribute-level attention layer assigns higher weights to key fields (such as numerical attributes). Qwen2, an open-source model developed by Alibaba Cloud, is a set of advanced basic and pre-tuned language models with parameters ranging from 50 million to 72 billion, including dense models and expert mixture models.

[0068] By performing relational encoding on the text description of the node, the interacting subjects in the relational description are extracted; the recurrent neural network (RNN) hidden layer of the language model is called, and combined with position encoding, the temporal dependency of the text description of the node (such as "years of cooperation") is captured to strengthen the relational description.

[0069] S452. Hierarchically extract edge semantic features of multiple levels of text description of the edge, fuse the edge semantic features of multiple levels, and obtain an edge semantic vector.

[0070] In a preferred implementation of this embodiment, an improved LLM encoding module is used to semantically encode the text description of the node and / or the text description of the edge; the LLM encoding module includes an input submodule, an LLM encoding submodule, an adaptation submodule, a hierarchical attention submodule and a knowledge enhancement submodule;

[0071] The text description of the node and / or the text description of the edge is input into the LLM encoding submodule through the input submodule. The LLM encoding submodule realizes the conversion from text to vector based on the Transformer layer of Qwen2. The adaptation submodule is a fully connected network. The adaptation submodule can map the general semantic vector to the feature space adapted to the dimension of GNN, and adapt the 256-dimensional vector adapted to the language large model LLM neural network GNN to enhance the compatibility of the downstream graph structure.

[0072] The hierarchical attention submodule includes an attribute-level attention layer and a relationship-level attention layer, which are added sequentially after the Transformer layer. The attribute-level attention layer can focus on key fields in the node attribute description (such as "total assets"), and the relationship-level attention layer is used to strengthen the interactive signals in the relationship description (such as "number of cooperative projects"), and weights the attribute keywords (such as numeric fields) and relationship descriptions (such as "cooperative projects") separately. The knowledge enhancement submodule is used to optimize the vector space distribution through contrastive learning (such as positive and negative sample pairs), incorporate professional knowledge in the field of enterprise risk, and use external textual knowledge to enhance the semantic understanding and relational reasoning capabilities of GNN. External textual knowledge refers to information extracted from resources outside the training data (such as knowledge graphs, dictionaries, and rule bases). Resources outside the training data include but are not limited to industry rule bases, domain dictionaries, etc.

[0073] Specifically, the positive sample pairs are different descriptive texts of the same enterprise (such as annual report data vs. news reports), and the negative sample pairs are similar descriptions of different enterprises (such as enterprises in the same industry but with different credit ratings). Comparison is used to shorten the distance between positive samples, push away negative samples, and optimize the rationality of the distribution of the semantic space.

[0074] For the extracted multi-level node semantic features (or edge semantic features), the node semantic features (or edge semantic features) of different layers are spliced ​​or weightedly fused.

[0075] With reference to the following table, in order to facilitate understanding by those skilled in the art, the function of the improved LLM encoding module of this embodiment is verified with reference to specific examples:

[0076]

[0077] After mapping by the adaptation layer, the embedding distance between the asset risk vector of enterprise A and similar risk enterprises is shortened (cosine similarity reaches 0.92), while it is farther away from risk-free enterprises (cosine similarity is 0.15), proving that the improved LLM encoding module effectively captures the risk semantic characteristics.

[0078] The calculation process of the cosine similarity between the asset risk vector of enterprise A and the vector corresponding to similar risk enterprises, as well as the cosine similarity between the asset risk vector of enterprise A and the vector corresponding to risk-free enterprises is as follows:

[0079] The calculation formula for cosine similarity is:

[0080]

[0081] Where, “·” represents dot product, A and B represent two sets of vectors, A·B=∑(A i *B i), i represents the vector dimension index, i∈[1,768];

[0082] ||.|| represents the modulus of the vector, and ||A|| represents the modulus of vector A; ||B|| represents the modulus of vector B, “*” indicates multiplication.

[0083] Assume that the embedding vector of enterprise A is [0.23, -0.45, 0.67, ...] (768 dimensions in total);

[0084] The embedding vector of similar venture enterprises is [0.25, -0.43, 0.65, ...] (768 dimensions in total);

[0085] The embedding vector of risk-free enterprises is [0.89, 0.12, -0.31, ...] (768 dimensions in total).

[0086] The calculation process of the cosine similarity sim1 between the asset risk vector of enterprise A and similar risk enterprises is as follows:

[0087]

[0088] The calculation process of the cosine similarity sim2 between the asset risk vector of enterprise A and the risk-free enterprise is as follows:

[0089]

[0090] S5. Update the representation of the nodes in the original enterprise knowledge graph according to the node semantic vector and the edge semantic vector to obtain a new enterprise knowledge graph.

[0091] Specifically, the representations of nodes and edges in the original enterprise knowledge graph are updated according to the node semantic vectors and edge semantic vectors to obtain a new enterprise knowledge graph, including:

[0092] S51. For each node in the original enterprise knowledge graph, aggregate the node semantic vectors of the nodes adjacent to the node to obtain a first fusion feature;

[0093] S52. Update the representation of each node according to the node semantic vector corresponding to each node and the first fusion feature to obtain a new enterprise knowledge graph.

[0094] Specifically, an adjacency matrix of a graph is constructed based on the enterprise relationships (such as shareholding, cooperation, and litigation relationships) in the original enterprise risk data. The adjacency matrix of the graph can be a directed or undirected graph.

[0095] For the node semantic vectors of nodes adjacent to the node, a splicing method or a weighted fusion method can be used; when performing weighted fusion, the weight of the node semantic vector can be adjusted according to the edge information; the edge weight in the graph object can be quantified based on the relationship strength. For example, when the shareholding ratio of the subsidiary reaches a threshold, the weight is set to 1, otherwise it is 0.5.

[0096] S6. Analyze the new enterprise knowledge graph to obtain risk analysis results of the original enterprise investment and financing data.

[0097] By restructuring the graph prediction task logic, traditional graph tasks such as node classification, connection prediction, and graph classification are uniformly transformed into prediction problems for "interest nodes." Task prediction encompasses node classification, connection prediction, and graph classification. By adding node prompts to describe new tasks (such as predicting "supply chain disruption risk"), graph convolutional models can directly infer without requiring additional training data.

[0098] The purpose of node classification is to convert the original single-node classification into the association prediction between the target node and its virtual "category node".

[0099] The purpose of connection prediction is to predict whether there is an edge between two nodes, which is converted into the prediction of relationship category nodes (for example, inserting a "cooperation relationship exists" node).

[0100] The purpose of graph classification is to focus the classification task of the entire graph on the prediction of global virtual "graph attribute nodes". For example, when judging the risk level of an enterprise, the model needs to integrate all node information to predict the risk label of the global node.

[0101] In order to achieve efficient prediction of different tasks by graph convolutional models, task-independent prompt nodes are introduced when constructing the text attribute graph (TAG). For example, nodes representing global attributes (such as the "high-risk enterprise graph" node) are added to the graph classification task. The model learns the semantic features of these nodes and associates them with the actual prediction target.

[0102] Specifically, the new enterprise knowledge graph is parsed to obtain risk analysis results of the original enterprise investment and financing data, including:

[0103] S61. Optimize the network parameters of the graph convolutional network using the updated node representation and external text knowledge to obtain an optimized graph convolutional model.

[0104] S62. Use the optimized graph convolution model to perform task prediction to determine the risk category of nodes in the new enterprise knowledge graph and whether there is fraudulent association on the edges, and obtain the risk analysis results of the original enterprise investment and financing data.

[0105] External textual knowledge refers to information extracted from resources outside of training data (such as knowledge graphs, dictionaries, and rule bases). It is used to supplement the implicit semantics or domain characteristics of the text. This external textual knowledge enhances the model's understanding of textual semantics and addresses the issues of data sparsity and insufficient context. Specifically, a task prediction module is constructed based on an optimized graph convolutional model to predict tasks of varying granularity, including node classification, link prediction, and graph classification. Through task-independent graph representation, the task prediction module can be trained in one scenario and seamlessly transferred to other scenarios, enabling cross-domain and cross-task learning of graph structure information.

[0106] In the preferred implementation of this embodiment, in order to enhance the adaptability of the model to multiple tasks, the task prediction module adopts a mechanism that combines dynamic prompt generation with self-supervision, wherein the dynamic prompt generation mechanism automatically generates task description text (such as "predict whether there is a controlling relationship between Company A and Company B") through the Qwen2 large model, and encodes the task description text into a prompt node embedding, which is input into the graph convolution model as part of the graph; while the self-supervision mechanism allows the graph convolution model to reconstruct missing information based on contextual relationships by randomly masking some edges or node attributes. In addition, a positive and negative sample comparison is performed on the embeddings of the same node under different tasks, forcing the model to learn task-invariant features, enhancing the semantic understanding and relational reasoning capabilities of GNN, and supporting the generation of explainable risk conduction paths; the mechanism that combines dynamic prompt generation with self-supervision eliminates the need for the graph convolution model to be retrained for different tasks, and only requires adjusting the prompt nodes to adapt to new scenarios, further improving the migration capabilities across data sets and tasks.

[0107] Based on the same inventive concept, an embodiment of the present invention provides an enterprise investment and financing risk analysis system.

[0108] The enterprise investment and financing risk analysis system described in the present invention can be installed in an electronic device. According to the functions to be implemented, the enterprise investment and financing risk analysis system includes an acquisition module, a parsing module, a knowledge graph construction module, a processing module, an update module, and an output module, wherein:

[0109] The acquisition module can obtain original enterprise investment and financing data;

[0110] The parsing module can parse the original enterprise investment and financing data to obtain the graph demand information of enterprise investment risks;

[0111] The knowledge graph construction module can construct the original enterprise knowledge graph based on the graph requirement information. The original enterprise knowledge graph includes several nodes and edges connecting the nodes. The nodes represent the keywords in the original enterprise investment and financing data, and the edges are used to represent the relationship between the nodes.

[0112] The processing module can map nodes and edges to text semantic space respectively to obtain node semantic vectors and edge semantic vectors;

[0113] The update module can update the representation of nodes in the original enterprise knowledge graph according to the node semantic vector and edge semantic vector to obtain a new enterprise knowledge graph;

[0114] The output module can parse the new enterprise knowledge graph and obtain risk analysis results for the original enterprise investment and financing data.

[0115] The module described in the present invention may also be referred to as a unit, which refers to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, and is stored in a memory of the electronic device.

[0116] The various variations and specific examples of the enterprise investment and financing risk analysis method provided in the above embodiment are also applicable to the enterprise investment and financing risk analysis system of this embodiment. Through the above detailed description of the enterprise investment and financing risk analysis method, those skilled in the art can clearly understand the implementation method of the enterprise investment and financing risk analysis system of this embodiment. For the sake of brevity of the specification, it will not be described in detail here.

[0117] This application also discloses an electronic device, such as Figure 4 FIG2 is a schematic diagram of the structure of an electronic device for an enterprise investment and financing risk analysis method according to an embodiment of the present invention. The electronic device may include at least one processor 10, a memory 11 communicatively coupled to the at least one processor, a communication bus 12, and a communication interface 13. The electronic device may also include a computer program stored in the memory 11 and executable on the processor 10, such as a method program for analyzing enterprise investment and financing risks.

[0118] In some embodiments, the processor 10 may be composed of an integrated circuit, for example, a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips. The processor 10 is the control core (Control Unit) of the electronic device, connecting the various components of the entire electronic device using various interfaces and lines, and executing or executing programs or modules stored in the memory 11 (such as a method for performing corporate investment and financing risk analysis, etc.), as well as calling data stored in the memory 11, to perform various functions of the electronic device and process data.

[0119] The memory 11 includes at least one type of readable storage medium, and the readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (for example, an SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 11 may be an internal storage unit of an electronic device, such as a mobile hard disk of the electronic device. In other embodiments, the memory 11 may also be an external storage device of an electronic device, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Furthermore, the memory 11 may also include both an internal storage unit and an external storage device of the electronic device. The memory 11 can not only be used to store application software and various types of data installed in the electronic device, such as the code of a method program for enterprise investment and financing risk analysis, but can also be used to temporarily store data that has been output or is to be output.

[0120] The communication bus 12 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.

[0121] The communication interface 13 is used for communication between the above-mentioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a display (Display), an input unit (such as a keyboard (Keyboard)), optionally, the user interface may also be a standard wired interface, a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, for displaying information processed in the electronic device and for displaying a visual user interface.

[0122] Figure 4 Only the electronic device with components is shown, and it can be understood by those skilled in the art that Figure 4The structure shown does not constitute a limitation of the electronic device, and may include fewer or more components than shown, or combine certain components, or arrange the components differently. For example, although not shown, the electronic device may also include a power supply (such as a battery) to power each component. Preferably, the power supply can be logically connected to at least one processor 10 through a power management device, so that functions such as charging management, discharging management, and power consumption management are implemented through the power management device. The power supply may also include one or more DC or AC power supplies, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0123] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0124] Furthermore, if the module / unit integrated into the electronic device is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile.

[0125] The present application provides a computer-readable storage medium, for example, including any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM). The computer-readable storage medium stores a computer program capable of being loaded by a processor and executing the enterprise investment and financing risk analysis method of the above-described embodiment.

[0126] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "example," "specific example," "one implementation," "a preferred implementation," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0127] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A method for analyzing enterprise investment and financing risks, characterized in that: The method comprises: Obtain original enterprise investment and financing data; Analyze the original enterprise investment and financing data to obtain the map demand information of enterprise investment risks; Construct an original enterprise knowledge graph based on graph demand information. The original enterprise knowledge graph includes several nodes and several edges connecting the nodes. The nodes represent keywords in the original enterprise investment and financing data, and the edges are used to represent the association relationship between the nodes. Map nodes and edges to text semantic space respectively to obtain node semantic vectors and edge semantic vectors; Update the representation of nodes in the original enterprise knowledge graph according to the node semantic vector and edge semantic vector to obtain a new enterprise knowledge graph; Analyze the new enterprise knowledge graph to obtain risk analysis results of the original enterprise investment and financing data.

2. The enterprise investment and financing risk analysis method according to claim 1, characterized in that: The original enterprise investment and financing data includes structured data and semi-structured data. Before parsing the original enterprise investment and financing data to obtain the enterprise investment risk map demand information, the method further includes: Unifying the data format of original enterprise investment and financing data, and standardizing the original enterprise investment and financing data after the unified data format to obtain standardized enterprise investment and financing data; Based on standardized enterprise investment and financing data, we can obtain the map demand information of enterprise investment risks.

3. The enterprise investment and financing risk analysis method according to claim 1, characterized in that: The steps of mapping nodes and edges to text semantic space to obtain node semantic vectors and edge semantic vectors include: Extract the attribute names and attribute values ​​of nodes from the original enterprise investment and financing data to obtain node attributes; Extract edge information from the original enterprise investment and financing data to obtain edge attributes; Convert node attributes into natural language text descriptions to obtain text descriptions of the nodes; Convert edge attributes into natural language text descriptions to obtain edge text descriptions; Semantic encoding is performed on the text description of the node and the text description of the edge respectively to obtain the node semantic vector and edge semantic vector.

4. The enterprise investment and financing risk analysis method according to claim 3, characterized in that: The semantic encoding of the text description of the node and the text description of the edge is performed respectively to obtain the node semantic vector and the edge semantic vector, including: Hierarchically extract multiple levels of node semantic features from the text description of the node, fuse the multiple levels of node semantic features to obtain the node semantic vector; The edge semantic features of multiple levels of the text description of the edge are extracted hierarchically, and the edge semantic features of multiple levels are fused to obtain the edge semantic vector.

5. The enterprise investment and financing risk analysis method according to any one of claims 1 to 4, characterized in that: The construction of the original enterprise knowledge graph based on graph demand information includes: Determine graph data information based on graph requirement information. Graph data information includes node attribute names, node attribute values, and edge information. Use graph convolutional networks to construct nodes and edges in the original enterprise knowledge graph into a graph object; The representation of the node is determined according to the attribute name, attribute value and edge information of each node in the graph object to obtain the original enterprise knowledge graph; the representation of the node includes the node's neighbor structure and the node's semantic information.

6. The enterprise investment and financing risk analysis method according to claim 5, characterized in that: The updating of the representations of nodes and edges in the original enterprise knowledge graph according to the node semantic vectors and the edge semantic vectors to obtain a new enterprise knowledge graph includes: For each node in the original enterprise knowledge graph, aggregate the node semantic vectors of the nodes adjacent to the node to obtain a first fusion feature; The representation of each node is updated according to the node semantic vector corresponding to each node and the first fusion feature to obtain a new enterprise knowledge graph.

7. The enterprise investment and financing risk analysis method according to claim 6, characterized in that: The analysis of the new enterprise knowledge graph to obtain risk analysis results of the original enterprise investment and financing data includes: The updated node representation and external text knowledge are used to optimize the network parameters of the graph convolutional network to obtain an optimized graph convolutional model; An optimized graph convolutional model is used for task prediction to determine the risk category of nodes in the new enterprise knowledge graph and whether there is fraudulent association on the edges, thereby obtaining risk analysis results for the original enterprise investment and financing data.

8. An enterprise investment and financing risk analysis system, used to implement the enterprise investment and financing risk analysis method according to any one of claims 1 to 7, characterized in that: include: Acquisition module, used to obtain original enterprise investment and financing data; The parsing module is used to parse the original enterprise investment and financing data to obtain the graph demand information of the enterprise investment risk; The knowledge graph construction module is used to construct the original enterprise knowledge graph based on the graph requirement information. The original enterprise knowledge graph includes several nodes and edges connecting the nodes. The nodes represent keywords in the original enterprise investment and financing data, and the edges are used to represent the association relationship between the nodes. A processing module is used to map nodes and edges to text semantic space to obtain node semantic vectors and edge semantic vectors respectively; An updating module is used to update the representation of nodes in the original enterprise knowledge graph according to the node semantic vector and the edge semantic vector to obtain a new enterprise knowledge graph; The output module is used to parse the new enterprise knowledge graph and obtain the risk analysis results of the original enterprise investment and financing data.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor (10); and, a memory (11) communicatively coupled to the at least one processor (10); Wherein, the memory (11) stores a computer program that can be executed by the at least one processor (10), and the computer program is executed by the at least one processor (10) so that the at least one processor (10) can execute the enterprise investment and financing risk analysis method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program; when the computer program is executed by a processor, the enterprise investment and financing risk analysis method according to any one of claims 1 to 7 is implemented.

Images