Financial risk early warning method and device based on knowledge graph

Abstract

The embodiment of the application provides a kind of financial risk early warning method and device based on knowledge graph, method includes: financial business data is carried out data preprocessing and is constructed to obtain financial knowledge graph;According to the local neighborhood information aggregation operation of pre-set relationship graph convolutional neural network model to the financial knowledge graph, determine the corresponding topological feature vector;According to the semantic feature extraction of pre-set language model to the financial knowledge graph, determine the corresponding semantic feature vector;Through multilayer perception depth neural network model, the topological feature vector and the semantic feature vector are carried out feature fusion, and according to the feature fusion result and pre-set risk score function determine corresponding financial risk;The application can efficiently, accurately, comprehensively carry out financial risk early warning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence knowledge graph, in particular to a financial risk early warning method and device based on knowledge graph. BACKGROUND

[0002] The existing financial loan early warning strategy is to formulate some rules to judge whether the current loaner faces the risk of non-standard use of financial funds or inability to repay according to the prior knowledge and regulations of the field by experts.

[0003] The inventor finds that with the increasing amount of financial data, the existing technology needs to consume a large amount of manpower and time cost, at the same time, the flow direction of loan funds becomes wider and deeper, and experts are difficult to quickly mine the implicit correlation generated in the process of loan fund flow, so it is difficult to judge whether the current financial loan exists a large risk. SUMMARY

[0004] In view of the problems in the prior art, the present application provides a financial risk early warning method and device based on knowledge graph, which can efficiently, accurately and comprehensively perform financial risk early warning.

[0005] In order to solve at least one of the above problems, the present application provides the following technical solutions:

[0006] In a first aspect, the present application provides a financial risk early warning method based on knowledge graph, comprising:

[0007] data preprocessing of financial business data and construction to obtain a financial knowledge graph;

[0008] performing local neighborhood information aggregation operation on the financial knowledge graph according to a preset relationship graph convolutional neural network model to determine a corresponding topological feature vector;

[0009] performing semantic feature extraction on the financial knowledge graph according to a preset language model to determine a corresponding semantic feature vector;

[0010] performing feature fusion on the topological feature vector and the semantic feature vector through a multilayer perception deep neural network model, and determining a corresponding financial risk according to the feature fusion result and a preset risk scoring function.

[0011] Further, the data preprocessing of financial business data and the construction to obtain a financial knowledge graph comprise:

[0012] classifying the financial business data after data cleaning according to business types to obtain semi-structured financial business data after the classification processing, and performing knowledge extraction on the financial business data according to a preset extraction rule to obtain financial business knowledge;

[0013] A financial knowledge graph is constructed based on the preset graph database, the financial business nodes in the financial business knowledge, and the relationships between the financial business nodes.

[0014] Furthermore, the step of performing local neighborhood information aggregation on the financial knowledge graph based on a preset relational graph convolutional neural network model to determine the corresponding topological feature vector includes:

[0015] Based on the preset relational graph convolutional neural network model, the relationship types and directed edge directions between financial business nodes in the financial knowledge graph, local neighborhood information aggregation operations are performed to obtain the topological feature vectors of each financial business node and its corresponding relationship.

[0016] Further, the step of extracting semantic features from the financial knowledge graph based on a preset language model to determine the corresponding semantic feature vector includes:

[0017] The financial business nodes and their relationships in the financial knowledge graph are mapped according to a preset self-attention model to construct corresponding word vectors;

[0018] The corresponding semantic feature vector is determined by multiplying the word vector with a set random word vector matrix.

[0019] Furthermore, after determining the corresponding semantic feature vector based on the product of the word vector and the set random word vector matrix, the method further includes:

[0020] The semantic feature vector is subjected to a loss value calculated based on a preset cross-entropy loss function. The random word vector matrix is ​​then trained and iterated based on the calculated loss value and a preset backpropagation algorithm to obtain the random word vector matrix after the parameter training and iteration.

[0021] Furthermore, the feature fusion of the topological feature vector and the semantic feature vector using a multilayer perceptual deep neural network model includes:

[0022] By performing linear operations on the topological feature vector and the semantic feature vector using a deep neural network multilayer perceptron, the topological feature vector and the semantic feature vector are mapped to the same vector space to obtain the feature fusion result.

[0023] Secondly, this application provides a knowledge graph-based financial risk early warning device, comprising:

[0024] The knowledge graph construction module is used to preprocess financial business data and construct a financial knowledge graph.

[0025] The topology feature determination module is used to perform local neighborhood information aggregation on the financial knowledge graph according to a preset relation graph convolutional neural network model to determine the corresponding topology feature vector.

[0026] The semantic feature determination module is used to extract semantic features from the financial knowledge graph according to a preset language model and determine the corresponding semantic feature vector.

[0027] The feature fusion risk assessment module is used to perform feature fusion on the topological feature vector and the semantic feature vector through a multilayer perceptual deep neural network model, and determine the corresponding financial risk based on the feature fusion result and a preset risk scoring function.

[0028] Furthermore, the knowledge graph construction module includes:

[0029] The knowledge extraction unit is used to classify the cleaned financial business data according to the business type to obtain semi-structured financial business data after the classification process, and to extract knowledge from the financial business data according to the preset extraction rules to obtain financial business knowledge.

[0030] The graph construction unit is used to construct a financial knowledge graph based on a preset graph database, financial business nodes in the financial business knowledge, and the relationships between the financial business nodes.

[0031] Furthermore, the topology feature determination module includes:

[0032] The local neighborhood information aggregation unit is used to perform local neighborhood information aggregation operations based on the preset relationship graph convolutional neural network model, the relationship types between financial business nodes in the financial knowledge graph, and the direction of directed edges, to obtain the topological feature vectors of each financial business node and its corresponding relationship.

[0033] Furthermore, the semantic feature determination module includes:

[0034] The word vector construction unit is used to map each financial business node and its relationship in the financial knowledge graph according to a preset self-attention model, and construct the corresponding word vectors.

[0035] The semantic feature extraction unit is used to determine the corresponding semantic feature vector based on the product of the word vector and a set random word vector matrix.

[0036] Furthermore, the semantic feature determination module also includes:

[0037] The loss training unit is used to calculate the loss value of the semantic feature vector according to the preset cross-entropy loss function, and to perform parameter training and parameter iteration on the random word vector matrix according to the calculated loss value and the preset backpropagation algorithm to obtain the random word vector matrix after parameter training and parameter iteration.

[0038] Furthermore, the feature fusion risk assessment module includes:

[0039] The feature fusion unit is used to perform linear operations on the topological feature vector and the semantic feature vector through a deep neural network multilayer perceptron, mapping the topological feature vector and the semantic feature vector to the same vector space to obtain the feature fusion result.

[0040] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the knowledge graph-based financial risk early warning method.

[0041] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the knowledge graph-based financial risk early warning method described above.

[0042] Fifthly, this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the knowledge graph-based financial risk early warning method.

[0043] As can be seen from the above technical solution, this application provides a financial risk early warning method and device based on knowledge graph. By comprehensively considering the topological feature vectors and semantic feature vectors in the knowledge graph, the accuracy of financial risk early warning is improved. Furthermore, the topological feature vectors and semantic feature vectors are fused using a multilayer perceptual deep neural network model, and the corresponding financial risk is determined based on the feature fusion result and a preset risk scoring function. Thus, financial risk early warning can be carried out efficiently, accurately, and comprehensively. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is one of the flowcharts illustrating the knowledge graph-based financial risk early warning method in the embodiments of this application;

[0046] Figure 2 This is the second flowchart illustrating the knowledge graph-based financial risk early warning method in this application embodiment;

[0047] Figure 3 This is the third flowchart illustrating the knowledge graph-based financial risk early warning method in this application embodiment;

[0048] Figure 4 This is one of the structural diagrams of the knowledge graph-based financial risk early warning device in the embodiments of this application;

[0049] Figure 5 This is the second structural diagram of the knowledge graph-based financial risk early warning device in the embodiments of this application;

[0050] Figure 6 This is the third structural diagram of the knowledge graph-based financial risk early warning device in the embodiments of this application;

[0051] Figure 7 This is the fourth structural diagram of the knowledge graph-based financial risk early warning device in the embodiments of this application;

[0052] Figure 8 This is the fifth structural diagram of the knowledge graph-based financial risk early warning device in the embodiments of this application;

[0053] Figure 9 This is the sixth structural diagram of the knowledge graph-based financial risk early warning device in the embodiments of this application;

[0054] Figure 10 This is a schematic diagram of the structure of the electronic device in the embodiments of this application. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0056] It is worth noting that the acquisition, storage, use, and processing of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations.

[0057] Considering that existing technologies require significant manpower and time, and that financial funds are flowing more broadly and deeply, making it difficult for experts to quickly uncover implicit relationships within these flows and thus assess the potential for substantial financial risks, this application provides a knowledge graph-based financial risk early warning method and apparatus. By comprehensively considering topological and semantic feature vectors within the knowledge graph, the accuracy of financial risk early warning is improved. Furthermore, a multi-layer perceptual deep neural network model is used to fuse the topological and semantic feature vectors, and the corresponding financial risk is determined based on the feature fusion result and a preset risk scoring function. This enables efficient, accurate, and comprehensive financial risk early warning.

[0058] To enable efficient, accurate, and comprehensive financial risk early warning, this application provides an embodiment of a knowledge graph-based financial risk early warning method. See [link to embodiment]. Figure 1 The knowledge graph-based financial risk early warning method specifically includes the following:

[0059] Step S101: Perform data preprocessing on financial business data and construct a financial knowledge graph.

[0060] Optionally, the financial business data mentioned in this application refers to common business data in the financial field, such as loan business data. Similarly, the financial knowledge graph can also be specifically a loan knowledge graph.

[0061] Optionally, the data preprocessing performed on the financial business data includes, but is not limited to, data cleaning, data classification, and knowledge extraction.

[0062] The data cleaning can be achieved using existing data cleaning techniques.

[0063] The data classification refers to classifying financial business data according to business type. For example, financial business data can be divided into various types such as financial customer information, financial product information, and financial contract information. These data are mainly structured data in relational databases.

[0064] It is understood that when the financial business data specifically refers to loan business data, the loan business data can be divided into various types of data such as loan customer information, loan product information, and loan contract information.

[0065] The knowledge extraction can be performed by obtaining semi-structured financial business data after the above data cleaning and / or data classification, and then extracting knowledge through preset extraction rules to obtain financial business knowledge.

[0066] It can be understood that the financial business knowledge is a triple-structured data used to construct a financial knowledge graph. A triple is the organization method of knowledge in the knowledge graph, usually represented in the structural form of <head entity, relationship, tail entity>. For example, in the triple of a specific loan business knowledge, <Company A, borrow, Bank B> represents the loan business knowledge that Company A borrows from Bank B.

[0067] Optionally, the present application can construct a financial knowledge graph by combining a preset graph database (such as Neo4J graph database) with the financial business nodes in the financial business knowledge and the association relationships between the financial business nodes.

[0068] It can be understood that the financial business nodes are the entities in the financial knowledge graph, used to represent financial business information, and are connected by a directed edge between each financial business node. The directed edge is respectively used to represent the association relationship between the financial business information corresponding to the two financial business nodes connected by it. The financial business information is the attributes contained in the entity (financial account, financial amount, credit record, transaction details, etc.), and the association relationship is the action or event that occurs between two entities, and its types mainly include (borrowing, repayment, guarantee, etc.).

[0069] Step S102: Perform a local neighborhood information aggregation operation on the financial knowledge graph according to a preset relational graph convolutional neural network model to determine the corresponding topological feature vector.

[0070] It can be understood that in the financial knowledge graph, a piece of financial knowledge represents a fact, usually represented in the form of a triple structure. In order to enable the subsequent model to understand and operate on the financial knowledge in the financial knowledge graph, the present application can convert the fact from natural language into a feature vector, and this process is usually also called embedding. In order to complete this process, the present application can construct a fact information encoder from a topological perspective to determine the topological feature vector of the financial knowledge graph (including entity embedding vector and relationship embedding vector).

[0071] Optionally, the present application can perform a local neighborhood information aggregation operation on the financial knowledge graph through a preset relational graph convolutional neural network model to determine the corresponding topological feature vector.

[0072] Optionally, the relational graph convolutional neural network model of the present application can perform a local neighborhood information aggregation operation according to the association relationship type and the directed edge direction between the financial business nodes in the financial knowledge graph to obtain the topological feature vectors of each financial business node and the corresponding association relationship. The specific aggregation function can be:

[0073]

[0074] Where R represents the set of relations, N represents the set of nodes, is the regularization term, W is the parameter of the current layer, h is the feature of the layer corresponding to the node, j represents the node in the set of neighboring nodes, i represents the number of layers of the current graph convolution, and r represents the relation in the set of relations.

[0075] Step S103: Extract semantic features from the financial knowledge graph according to the preset language model and determine the corresponding semantic feature vector.

[0076] It is understandable that financial knowledge graphs contain a large amount of textual information, such as entity names, descriptions, and attributes. This information is of great significance for financial risk early warning. For example, when the real estate industry faces a huge crisis, all nodes whose names, attributes, or descriptions are related to real estate may face certain risks. Therefore, this application can use a pre-set language model (such as the KG-Bert language model) to extract the semantic features of nodes.

[0077] Optionally, the KG-Bert language model, targeting the triple structure in financial knowledge graphs, can treat the head entity, relation, and tail entity as the subject, predicate, and object of a sentence. Furthermore, this application can use a pre-defined self-attention model (such as the Transformer self-attention model) to map entities and relations into word vectors, thereby capturing the semantic features of the entire triple. The word vectors are used to call and train the model.

[0078] Optionally, this application can map each financial business node and its relationship in the financial knowledge graph according to a preset self-attention model to construct the corresponding word vectors, and then determine the corresponding semantic feature vectors by multiplying them with a set random word vector matrix. The specific steps can be as follows:

[0079] (1) Constructing a vocabulary list

[0080] (2) Construct one-hot vectors (i.e., word vectors)

[0081] (3) Construct a word vector matrix (numerical randomization)

[0082] (4) The word vector (i.e., semantic feature vector) corresponding to the one-hot vector can be obtained by multiplying the one-hot vector with the word vector matrix.

[0083] Step S104: Perform feature fusion on the topological feature vector and the semantic feature vector using a multilayer perceptual deep neural network model, and determine the corresponding financial risk based on the feature fusion result and a preset risk scoring function.

[0084] It is understandable that a multilayer perceptron is a deep neural network composed of parameter matrices. Since the vector spaces of topological feature vectors and semantic feature vectors are different, this application can perform linear operations on them using a multilayer perceptron to map these two features into the same vector space, thereby achieving feature fusion. The corresponding financial risk is then determined based on the feature fusion result and a preset risk scoring function.

[0085] Optionally, the risk scoring function described in this application may be, for example, the Softmax function, which is essentially an activation function that can map the outputs of multiple neurons to the (0, 1) interval and ensure that the sum of all their outputs is 1. The output of this risk scoring function can be regarded as a probability, thereby performing multi-classification tasks.

[0086] The formula for the Softmax function is as follows:

[0087]

[0088] The final output is a vector S, where the value of each dimension represents the amount of information about its corresponding category. The Softmax function compares the exponent of each dimension with the sum of the exponents of all elements to calculate the probability of its corresponding category. This can be used as the financial risk score of the current financial knowledge graph (i.e., financial business data) in this application. The corresponding financial risk can be determined by comparing the score with the score threshold.

[0089] As can be seen from the above description, the knowledge graph-based financial risk early warning method provided in this application can improve the accuracy of financial risk early warning by comprehensively considering the topological feature vectors and semantic feature vectors in the knowledge graph. Furthermore, it can perform feature fusion on the topological feature vectors and semantic feature vectors through a multi-layer perceptual deep neural network model, and determine the corresponding financial risk based on the feature fusion result and a preset risk scoring function. Thus, it can perform financial risk early warning efficiently, accurately and comprehensively.

[0090] In order to accurately construct a financial knowledge graph, in one embodiment of the knowledge graph-based financial risk early warning method of this application, see [link to relevant documentation]. Figure 2 The above step S101 may also specifically include the following:

[0091] Step S201: The cleaned financial business data is classified according to business type to obtain semi-structured financial business data after classification. Knowledge is extracted from the financial business data according to preset extraction rules to obtain financial business knowledge.

[0092] Step S202: Construct a financial knowledge graph based on a preset graph database, financial business nodes in the financial business knowledge, and the association relationships between the financial business nodes.

[0093] Optionally, the data preprocessing performed on the financial business data includes, but is not limited to, data cleaning, data classification, and knowledge extraction.

[0094] Among them, the data cleaning can be implemented using data cleaning techniques in the prior art.

[0095] The data classification refers to classifying financial business data according to business types. For example, financial business data is classified into various types of data such as financial customer information, financial product information, and financial contract information. These data are mainly structured data in a relational database.

[0096] The knowledge extraction can obtain semi-structured financial business data after the above data cleaning and / or data classification, and then perform knowledge extraction through preset extraction rules to obtain financial business knowledge.

[0097] It can be understood that the financial business knowledge is a triple structure data used to constitute the financial knowledge graph. A triple is the organization method of knowledge in the knowledge graph, usually represented in the structure form of <head entity, relationship, tail entity>. For example, in the triple of a specific loan business knowledge, <Company A, borrow, Bank B> represents the loan business knowledge that Company A borrows from Bank B.

[0098] Optionally, in this application, a financial knowledge graph can be constructed by combining a preset graph database (such as the Neo4J graph database) with the financial business nodes in the financial business knowledge and the association relationships between the financial business nodes.

[0099] It can be understood that the financial business nodes are the entities in the financial knowledge graph, used to represent financial business information, and are connected by a directed edge between each financial business node. The directed edges are respectively used to represent the association relationships between the financial business information corresponding to the two financial business nodes connected by them.

[0100] In order to accurately obtain the topological feature vectors in the knowledge graph, in an embodiment of the financial risk warning method based on the knowledge graph in this application, the above step S102 may further specifically include the following content:

[0101] Perform a local neighborhood information aggregation operation according to a preset relational graph convolutional neural network model, the association relationship types and directed edge directions between the financial business nodes in the financial knowledge graph, to obtain the topological feature vectors of each financial business node and the corresponding association relationships.

[0102] Optionally, the relational graph convolutional neural network model of this application can perform local neighborhood information aggregation operations based on the relationship type and directed edge direction between each financial business node in the financial knowledge graph to obtain the topological feature vector of each financial business node and its corresponding relationship. The specific aggregation function can be:

[0103]

[0104] Where R represents the set of relations, N represents the set of nodes, is the regularization term, W is the parameter of the current layer, h is the feature of the layer corresponding to the node, j represents the node in the set of neighboring nodes, i represents the number of layers of the current graph convolution, and r represents the relation in the set of relations.

[0105] In order to accurately obtain the semantic feature vectors in the knowledge graph, in one embodiment of the knowledge graph-based financial risk early warning method of this application, see [link to relevant documentation]. Figure 3 The above step S103 may also specifically include the following:

[0106] Step S301: Map each financial business node and its relationship in the financial knowledge graph according to the preset self-attention model to construct the corresponding word vector.

[0107] Step S302: Determine the corresponding semantic feature vector based on the product of the word vector and the set random word vector matrix.

[0108] Optionally, this application can use a pre-defined language model (e.g., the KG-Bert language model) to extract the semantic features of nodes. The KG-Bert language model, designed for the triple structure in financial knowledge graphs, can treat the head entity, relation, and tail entity as the subject, predicate, and object of a sentence. Furthermore, this application can use a pre-defined self-attention model (e.g., the Transformer self-attention model) to map entities and relations into word vectors, thereby capturing the semantic features of the entire triple. These word vectors are used to invoke and train the model.

[0109] Optionally, this application can map each financial business node and its relationship in the financial knowledge graph according to a preset self-attention model to construct the corresponding word vectors, and then determine the corresponding semantic feature vectors by multiplying them with a set random word vector matrix. The specific steps can be as follows:

[0110] (1) Constructing a vocabulary list

[0111] (2) Construct one-hot vectors (i.e., word vectors)

[0112] (3) Construct a word vector matrix (numerical randomization)

[0113] (4) The word vector (i.e., semantic feature vector) corresponding to the one-hot vector can be obtained by multiplying the one-hot vector with the word vector matrix.

[0114] In order to improve the accuracy of semantic extraction, in one embodiment of the knowledge graph-based financial risk early warning method of this application, after step S302 above, the following content may be specifically included:

[0115] The semantic feature vector is subjected to a loss value calculated based on a preset cross-entropy loss function. The random word vector matrix is ​​then trained and iterated based on the calculated loss value and a preset backpropagation algorithm to obtain the random word vector matrix after the parameter training and iteration.

[0116] Optionally, since the word vector matrix is ​​initially a random value, this application can calculate the loss value of the semantic feature vector according to a preset cross-entropy loss function, and backpropagate it back to the random word vector matrix for parameter training and parameter iteration based on the training results, so as to make the random word vector matrix of this application more accurate.

[0117] In order to accurately perform feature fusion, in one embodiment of the knowledge graph-based financial risk early warning method of this application, the above step S103 may further include the following:

[0118] By performing linear operations on the topological feature vector and the semantic feature vector using a deep neural network multilayer perceptron, the topological feature vector and the semantic feature vector are mapped to the same vector space to obtain the feature fusion result.

[0119] It is understandable that a multilayer perceptron is a deep neural network composed of parameter matrices. Since the vector spaces of topological feature vectors and semantic feature vectors are different, this application can perform linear operations on them using a multilayer perceptron to map these two features into the same vector space, thereby achieving feature fusion. The corresponding financial risk is then determined based on the feature fusion result and a preset risk scoring function.

[0120] To enable efficient, accurate, and comprehensive financial risk early warning, this application provides an embodiment of a knowledge graph-based financial risk early warning device for implementing all or part of the aforementioned knowledge graph-based financial risk early warning method. See [link to embodiment]. Figure 4 The knowledge graph-based financial risk early warning device specifically includes the following components:

[0121] The knowledge graph construction module 10 is used to preprocess financial business data and construct a financial knowledge graph.

[0122] The topology feature determination module 20 is used to perform local neighborhood information aggregation operation on the financial knowledge graph according to the preset relationship graph convolutional neural network model to determine the corresponding topology feature vector.

[0123] The semantic feature determination module 30 is used to extract semantic features from the financial knowledge graph according to a preset language model and determine the corresponding semantic feature vector.

[0124] The feature fusion risk assessment module 40 is used to perform feature fusion on the topological feature vector and the semantic feature vector through a multilayer perceptual deep neural network model, and determine the corresponding financial risk based on the feature fusion result and a preset risk scoring function.

[0125] As can be seen from the above description, the knowledge graph-based financial risk early warning device provided in this application embodiment can improve the accuracy of financial risk early warning by comprehensively considering the topological feature vectors and semantic feature vectors in the knowledge graph. Furthermore, it can perform feature fusion on the topological feature vectors and semantic feature vectors through a multi-layer perceptual deep neural network model, and determine the corresponding financial risk based on the feature fusion result and a preset risk scoring function. Thus, it can perform financial risk early warning efficiently, accurately, and comprehensively.

[0126] In order to accurately construct a financial knowledge graph, in one embodiment of the knowledge graph-based financial risk early warning device of this application, see [link to relevant documentation]. Figure 5 The knowledge graph construction module 10 includes:

[0127] The knowledge extraction unit 11 is used to classify the cleaned financial business data according to the business type to obtain semi-structured financial business data after the classification process, and to extract knowledge from the financial business data according to the preset extraction rules to obtain financial business knowledge.

[0128] The graph construction unit 12 is used to construct a financial knowledge graph based on a preset graph database, financial business nodes in the financial business knowledge, and the relationships between the financial business nodes.

[0129] In order to accurately obtain the topological feature vectors in the knowledge graph, in one embodiment of the knowledge graph-based financial risk early warning device of this application, see [reference needed]. Figure 6 The topology feature determination module 20 includes:

[0130] The local neighborhood information aggregation unit 21 is used to perform local neighborhood information aggregation operations based on the preset relationship graph convolutional neural network model, the relationship type between each financial business node in the financial knowledge graph and the direction of the directed edge, to obtain the topological feature vector of each financial business node and the corresponding relationship.

[0131] In order to accurately obtain the semantic feature vectors in the knowledge graph, in one embodiment of the knowledge graph-based financial risk early warning device of this application, see [reference needed]. Figure 7 The semantic feature determination module 30 includes:

[0132] The word vector construction unit 31 is used to map each financial business node and its relationship in the financial knowledge graph according to a preset self-attention model, and construct the corresponding word vector.

[0133] The semantic feature extraction unit 32 is used to determine the corresponding semantic feature vector based on the product of the word vector and the set random word vector matrix.

[0134] To improve the accuracy of semantic extraction, in one embodiment of the knowledge graph-based financial risk early warning device of this application, see [link to relevant documentation]. Figure 8 The semantic feature determination module 30 further includes:

[0135] The loss training unit 33 is used to calculate the loss value of the semantic feature vector according to the preset cross-entropy loss function, and to perform parameter training and parameter iteration on the random word vector matrix according to the calculated loss value and the preset backpropagation algorithm to obtain the random word vector matrix after parameter training and parameter iteration.

[0136] To enable accurate feature fusion, in one embodiment of the knowledge graph-based financial risk early warning device of this application, see [link to relevant documentation]. Figure 9 The feature fusion risk assessment module 40 includes:

[0137] The feature fusion unit 41 is used to perform linear operations on the topological feature vector and the semantic feature vector through a deep neural network multilayer perceptron, and map the topological feature vector and the semantic feature vector to the same vector space to obtain the feature fusion result.

[0138] From a hardware perspective, in order to efficiently, accurately, and comprehensively conduct financial risk early warning, this application provides an embodiment of an electronic device for implementing all or part of the knowledge graph-based financial risk early warning method, wherein the electronic device specifically includes the following:

[0139] The system comprises a processor, memory, a communications interface, and a bus; wherein the processor, memory, and communications interface communicate with each other via the bus; the communications interface is used to realize information transmission between the knowledge graph-based financial risk early warning device and core business systems, user terminals, and related databases and other related devices; the logic controller can be a desktop computer, tablet computer, or mobile terminal, etc., and this embodiment is not limited to these. In this embodiment, the logic controller can be implemented with reference to the embodiments of the knowledge graph-based financial risk early warning method and the knowledge graph-based financial risk early warning device in the embodiments, the content of which is incorporated herein, and repeated details will not be described again.

[0140] It is understood that the user terminal may include smartphones, tablet computers, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), in-vehicle devices, smart wearable devices, etc. Among these, the smart wearable devices may include smart glasses, smartwatches, smart bracelets, etc.

[0141] In practical applications, parts of the knowledge graph-based financial risk early warning method can be executed on the electronic device side as described above, or all operations can be completed on the client device. The choice can be made based on the processing power of the client device and the limitations of the user's usage scenario. This application does not impose any limitations on this. If all operations are completed on the client device, the client device may further include a processor.

[0142] The aforementioned client device may have a communication module (i.e., a communication unit) that can communicate with a remote server to achieve data transmission. The server may include a server on the task scheduling center side; in other implementation scenarios, it may also include a server on an intermediate platform, such as a server on a third-party server platform that has a communication link with the task scheduling center server. The server may include a single computer device, a server cluster consisting of multiple servers, or a distributed server structure.

[0143] Figure 10 This is a schematic block diagram illustrating the system configuration of the electronic device 9600 according to an embodiment of this application. Figure 10 As shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It is worth noting that... Figure 10 This is an example; other types of structures can also be used to supplement or replace this structure to achieve telecommunications functions or other functions.

[0144] In one embodiment, the knowledge graph-based financial risk early warning method can be integrated into the central processing unit 9100. The central processing unit 9100 can be configured to perform the following controls:

[0145] Step S101: Perform data preprocessing on financial business data and construct a financial knowledge graph.

[0146] Step S102: Perform local neighborhood information aggregation on the financial knowledge graph according to the preset relation graph convolutional neural network model to determine the corresponding topological feature vector.

[0147] Step S103: Extract semantic features from the financial knowledge graph according to the preset language model and determine the corresponding semantic feature vector.

[0148] Step S104: Perform feature fusion on the topological feature vector and the semantic feature vector using a multilayer perceptual deep neural network model, and determine the corresponding financial risk based on the feature fusion result and a preset risk scoring function.

[0149] As can be seen from the above description, the electronic device provided in this application improves the accuracy of financial risk warning by comprehensively considering the topological feature vectors and semantic feature vectors in the knowledge graph. Furthermore, it performs feature fusion on the topological feature vectors and semantic feature vectors through a multi-layer perceptual deep neural network model, and determines the corresponding financial risk based on the feature fusion result and a preset risk scoring function. Thus, it can perform financial risk warning efficiently, accurately, and comprehensively.

[0150] In another embodiment, the knowledge graph-based financial risk early warning device can be configured separately from the central processing unit 9100. For example, the knowledge graph-based financial risk early warning device can be configured as a chip connected to the central processing unit 9100, and the knowledge graph-based financial risk early warning method function can be realized through the control of the central processing unit.

[0151] like Figure 10 As shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It is worth noting that the electronic device 9600 does not necessarily need to include these components. Figure 10 All components shown; in addition, the electronic device 9600 may also include Figure 10 For components not shown, please refer to existing technologies.

[0152] like Figure 10As shown, the central processing unit 9100, sometimes also referred to as a controller or operating control, may include a microprocessor or other processor device and / or logic device, which receives inputs and controls the operation of various components of the electronic device 9600.

[0153] The memory 9140 may be, for example, one or more of a cache, flash memory, hard drive, removable media, volatile memory, non-volatile memory, or other suitable devices. It may store the aforementioned failure-related information, and also store a program for executing that information. The central processing unit 9100 may execute the program stored in the memory 9140 to perform information storage or processing, etc.

[0154] Input unit 9120 provides input to central processing unit 9100. Input unit 9120 may be, for example, a keypad or touch input device. Power supply 9170 provides power to electronic device 9600. Display 9160 displays images and text. Display may be, for example, an LCD display, but is not limited thereto.

[0155] The memory 9140 can be a solid-state memory, such as a read-only memory (ROM), random access memory (RAM), a SIM card, etc. It can also be a memory that retains information even when power is off, can be selectively erased, and contains more data; examples of this type of memory are sometimes referred to as EPROMs. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application / function storage unit 9142 for storing application programs and function programs or processes for executing the operation of the electronic device 9600 via the central processing unit 9100.

[0156] The memory 9140 may also include a data storage unit 9143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 may include various drivers for the electronic device's communication functions and / or for performing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0157] The communication module 9110 is a transmitter / receiver 9110 that transmits and receives signals via the antenna 9111. The communication module (transmitter / receiver) 9110 is coupled to the central processing unit 9100 to provide input signals and receive output signals, which can be the same as in a conventional mobile communication terminal.

[0158] Based on different communication technologies, multiple communication modules 9110 can be configured in the same electronic device, such as cellular network modules, Bluetooth modules, and / or wireless LAN modules. The communication module (transmitter / receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby realizing typical telecommunications functions. The audio processor 9130 may include any suitable buffer, decoder, amplifier, etc. Additionally, the audio processor 9130 is coupled to a central processing unit 9100, enabling on-device recording via the microphone 9132 and on-device playback of stored sound via the speaker 9131.

[0159] Embodiments of this application also provide a computer-readable storage medium capable of implementing all steps of the knowledge graph-based financial risk early warning method with a server or client as the execution subject in the above embodiments. The computer-readable storage medium stores a computer program that, when executed by a processor, implements all steps of the knowledge graph-based financial risk early warning method with a server or client as the execution subject in the above embodiments. For example, when the processor executes the computer program, it implements the following steps:

[0160] Step S101: Perform data preprocessing on financial business data and construct a financial knowledge graph.

[0161] Step S102: Perform local neighborhood information aggregation on the financial knowledge graph according to the preset relation graph convolutional neural network model to determine the corresponding topological feature vector.

[0162] Step S103: Extract semantic features from the financial knowledge graph according to the preset language model and determine the corresponding semantic feature vector.

[0163] Step S104: Perform feature fusion on the topological feature vector and the semantic feature vector using a multilayer perceptual deep neural network model, and determine the corresponding financial risk based on the feature fusion result and a preset risk scoring function.

[0164] As can be seen from the above description, the computer-readable storage medium provided in this application improves the accuracy of financial risk warning by comprehensively considering the topological feature vectors and semantic feature vectors in the knowledge graph. Furthermore, it performs feature fusion on the topological feature vectors and semantic feature vectors through a multi-layer perceptual deep neural network model, and determines the corresponding financial risk based on the feature fusion result and a preset risk scoring function. Thus, it can efficiently, accurately and comprehensively provide financial risk warning.

[0165] Embodiments of this application also provide a computer program product capable of implementing all steps of the knowledge graph-based financial risk early warning method in the above embodiments, where the execution subject is a server or a client. When executed by a processor, this computer program / instruction implements the steps of the knowledge graph-based financial risk early warning method. For example, the computer program / instruction implements the following steps:

[0166] Step S101: Perform data preprocessing on financial business data and construct a financial knowledge graph.

[0167] Step S102: Perform local neighborhood information aggregation on the financial knowledge graph according to the preset relation graph convolutional neural network model to determine the corresponding topological feature vector.

[0168] Step S103: Extract semantic features from the financial knowledge graph according to the preset language model and determine the corresponding semantic feature vector.

[0169] Step S104: Perform feature fusion on the topological feature vector and the semantic feature vector using a multilayer perceptual deep neural network model, and determine the corresponding financial risk based on the feature fusion result and a preset risk scoring function.

[0170] As can be seen from the above description, the computer program product provided in this application improves the accuracy of financial risk warning by comprehensively considering the topological feature vectors and semantic feature vectors in the knowledge graph. Furthermore, it performs feature fusion on the topological feature vectors and semantic feature vectors through a multi-layer perceptual deep neural network model, and determines the corresponding financial risk based on the feature fusion result and a preset risk scoring function. Thus, it can efficiently, accurately and comprehensively provide financial risk warning.

[0171] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0172] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0173] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0174] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0175] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A financial risk early warning method based on knowledge graphs, characterized in that, The method includes: Financial business data is preprocessed and a financial knowledge graph is constructed. Based on a pre-defined relational graph convolutional neural network model, the financial knowledge graph is subjected to local neighborhood information aggregation to determine the corresponding topological feature vector; Semantic features are extracted from the financial knowledge graph based on a preset language model to determine the corresponding semantic feature vectors. A multilayer perceptual deep neural network model is used to perform linear operations on the topological feature vector and the semantic feature vector, mapping the topological feature vector and the semantic feature vector to the same vector space, thereby achieving feature fusion. The corresponding financial risk is then determined based on the feature fusion result and a preset risk scoring function. The step of performing local neighborhood information aggregation on the financial knowledge graph based on a preset relational graph convolutional neural network model to determine the corresponding topological feature vector includes: Based on the pre-defined relational graph convolutional neural network model, the relationship types between financial business nodes in the financial knowledge graph, and the direction of directed edges, a local neighborhood information aggregation operation is performed to obtain the topological feature vectors of each financial business node and its corresponding relationship. The relationship types include: borrowing, repayment, and guarantee. The step of extracting semantic features from the financial knowledge graph based on a preset language model to determine the corresponding semantic feature vector includes: The financial business nodes and their relationships in the financial knowledge graph are mapped according to a preset self-attention model to construct corresponding word vectors; The corresponding semantic feature vector is determined by multiplying the word vector with a set random word vector matrix; After determining the corresponding semantic feature vector based on the product of the word vector and the set random word vector matrix, the method further includes: The semantic feature vector is subjected to a loss value calculated based on a preset cross-entropy loss function. The random word vector matrix is ​​then subjected to parameter training and parameter iteration based on the calculated loss value and a preset backpropagation algorithm to obtain the random word vector matrix after parameter training and parameter iteration. The preset risk scoring function is the Softmax function formula: The Softmax function calculates the probability of a category by comparing the exponent of each dimension's value with the sum of the exponents of all element values. This probability is then compared with a scoring threshold to determine the corresponding financial risk.

2. The knowledge graph-based financial risk early warning method according to claim 1, characterized in that, The process of preprocessing financial business data and constructing a financial knowledge graph includes: After data cleaning, the financial business data is classified according to business type to obtain semi-structured financial business data after classification. Then, knowledge is extracted from the financial business data according to preset extraction rules to obtain financial business knowledge. A financial knowledge graph is constructed based on the preset graph database, the financial business nodes in the financial business knowledge, and the relationships between the financial business nodes.

3. A knowledge graph-based financial risk early warning device, characterized in that, include: The knowledge graph construction module is used to preprocess financial business data and construct a financial knowledge graph. The topology feature determination module is used to perform local neighborhood information aggregation on the financial knowledge graph according to a preset relation graph convolutional neural network model to determine the corresponding topology feature vector. The semantic feature determination module is used to extract semantic features from the financial knowledge graph according to a preset language model and determine the corresponding semantic feature vector. The feature fusion risk assessment module is used to perform linear operations on the topological feature vector and the semantic feature vector through a multilayer perceptual deep neural network model, mapping the topological feature vector and the semantic feature vector to the same vector space, thereby realizing feature fusion, and determining the corresponding financial risk based on the feature fusion result and a preset risk scoring function. The topology feature determination module includes: The local neighborhood information aggregation unit is used to perform local neighborhood information aggregation operations based on the preset relation graph convolutional neural network model, the relationship types between financial business nodes in the financial knowledge graph, and the direction of directed edges, to obtain the topological feature vectors of each financial business node and its corresponding relationship. The relationship types include: loan, repayment, and guarantee. The semantic feature determination module includes: The word vector construction unit is used to map each financial business node and its relationship in the financial knowledge graph according to a preset self-attention model, and construct the corresponding word vectors. The semantic feature extraction unit is used to determine the corresponding semantic feature vector based on the product of the word vector and a set random word vector matrix; The semantic feature determination module further includes: The loss training unit is used to calculate the loss value of the semantic feature vector according to the preset cross-entropy loss function, and to perform parameter training and parameter iteration on the random word vector matrix according to the calculated loss value and the preset backpropagation algorithm to obtain the random word vector matrix after parameter training and parameter iteration. The preset risk scoring function is the Softmax function formula: The Softmax function calculates the probability of a category by comparing the exponent of each dimension's value with the sum of the exponents of all element values. This probability is then compared with a scoring threshold to determine the corresponding financial risk.

4. The knowledge graph-based financial risk early warning device according to claim 3, characterized in that, The knowledge graph construction module includes: The knowledge extraction unit is used to classify the cleaned financial business data according to the business type to obtain semi-structured financial business data after the classification process, and to extract knowledge from the financial business data according to the preset extraction rules to obtain financial business knowledge. The graph construction unit is used to construct a financial knowledge graph based on a preset graph database, financial business nodes in the financial business knowledge, and the relationships between the financial business nodes.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the knowledge graph-based financial risk early warning method as described in claim 1 or 2.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the knowledge graph-based financial risk early warning method as described in claim 1 or 2.

7. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the knowledge graph-based financial risk early warning method as described in claim 1 or 2.

Images