Financial risk assessment method based on big data

Abstract

The invention discloses a financial risk assessment method based on big data, and relates to the technical field of finance, and the method comprises the following steps: S1, obtaining structured data, unstructured data and real-time streaming data of a target entity through a multi-source heterogeneous data collection module; s2, constructing an association relationship graph, and modeling risk propagation paths of a target entity and associated nodes thereof based on a graph neural network; s3, performing feature alignment and joint representation learning on the structured data, the unstructured text data and the time series data through a multi-modal data fusion module; and S4, based on the causal inference model, separating causal features and hybrid variables of the target entity risk event, generating causal risk factors, quantifying risk infection paths between nodes by setting an enterprise guarantee network and a supply chain relation graph dynamically constructed in a graph neural network, effectively identifying hidden risk nodes, and improving the risk assessment efficiency. And the chain reaction risk caused by the default of the associated enterprise is reduced.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular to a financial risk assessment method based on big data. Background Art

[0002] Under the wave of digitalization, big data technology has been deeply integrated into the financial risk assessment system, achieving revolutionary upgrades in risk identification, quantification and management through multi-dimensional data integration and intelligent algorithm models.

[0003] According to the patent title: Financial Risk Prediction Method and System Based on Big Data (patent publication number: CN118941391A, patent publication date: 2024-11-12), by utilizing big data technology and algorithms, various data are collected from the financial market, including market conditions, transactions, economic indicators, public opinion and other information, and through data collation, cleaning and exploratory analysis, the relationship, trend and impact of the data are evaluated; real-time analysis is conducted based on the characteristics of the financial market, and the trend correlation between the data and the impact on the market are comprehensively considered. The quality value of the data is evaluated and classified and marked. Finally, by analyzing the changes in the quality value of data over a period of time, the possible risk level of the financial market is predicted and warned in advance.

[0004] Based on the above-mentioned existing technologies, the current financial risk assessment methods based on big data still have the following problems. The existing methods rely on single entity data, ignore the correlation between equity relations, supply chain transactions, guarantee networks, etc. between enterprises, and it is difficult to quantify the risk propagation path in a complex network. For this reason, the present invention provides a financial risk assessment method based on big data. Summary of the Invention

[0005] In response to the shortcomings of existing technologies, the present invention provides a financial risk assessment method based on big data, which solves the following problems of existing financial risk assessment methods based on big data: existing methods rely on single entity data, ignore the correlation between equity relationships, supply chain transactions, guarantee networks, etc. between enterprises, and it is difficult to quantify the risk propagation path in complex networks.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a financial risk assessment method based on big data, comprising the following steps:

[0007] S1: Obtain the structured data, unstructured data and real-time streaming data of the target entity through the multi-source heterogeneous data acquisition module;

[0008] S2: Construct an association relationship graph and model the risk propagation path of the target entity and its associated nodes based on a graph neural network;

[0009] S3: Through the multimodal data fusion module, feature alignment and joint representation learning are performed on structured data, unstructured text data, and time series data;

[0010] S4: Based on the causal inference model, separate the causal characteristics and confounding variables of the target entity risk event and generate causal risk factors;

[0011] S5: Input the graph risk propagation weight, multimodal feature vector and causal risk factor into the risk assessment model and output the comprehensive risk score of the target entity;

[0012] S6: Trigger early warnings or generate dynamic risk control strategies based on risk scores.

[0013] Preferably, the application of the graph neural network in S2 includes: constructing a dynamic graph through corporate equity relations, supply chain transaction records and guarantee networks, using a graph attention network to calculate the risk contagion weights between nodes, and identifying hidden risk propagation paths based on a random walk algorithm.

[0014] Preferably, the implementation of the multimodal data fusion module in S3 includes: extracting time series features from structured data, performing semantic vectorization on unstructured data using the BERT model, aligning heterogeneous features through a cross-modal Transformer model, and generating a joint embedding representation.

[0015] Preferably, dual machine learning is used to eliminate confounding biases in observational data, counterfactual risk scenarios are constructed based on structural causal models, the causal effects of external shocks on target entities are quantified, and key causal chains are identified through causal discovery algorithms.

[0016] Preferably, the risk assessment model of S5 is an integrated model, including: taking the graph risk propagation weight as the input of the graph convolutional network, inputting the multimodal feature vector into the deep neural network branch, and making a joint decision on the causal risk factor and the above branch output through a weighted fusion layer.

[0017] Preferably, the data collection module in S1 further includes: obtaining pledge status data in real time through Internet of Things devices, and performing cross-institutional joint modeling based on a federated learning framework to protect data privacy.

[0018] Preferably, the S5 also includes a risk explainability module: using the SHAP algorithm to perform attribution analysis on the risk assessment results, generating a key risk feature contribution report, and visually displaying the risk transmission path and causal chain through a knowledge graph.

[0019] Preferably, the dynamic risk control strategy of S6 includes: adjusting the credit limit or margin ratio in real time based on reinforcement learning, and generating a stress test simulation report for systemic risk scenarios.

[0020] This invention provides a financial risk assessment method based on big data. Compared with the existing technology, it has the following advantages:

[0021] 1. This big data-based financial risk assessment method quantifies the risk transmission path between nodes by setting up a corporate guarantee network and supply chain relationship map dynamically constructed in a graph neural network, effectively identifies hidden risk nodes, and reduces the chain reaction risk caused by default of related companies.

[0022] 2. This big data-based financial risk assessment method, by setting up multimodal data fusion, integrates multi-source heterogeneous data such as financial statements, public opinion texts, and logistics trajectories, captures risk signals ignored by traditional models, and improves the risk assessment coverage of small and micro enterprises and entities with weak credit records.

[0023] 3. This big data-based financial risk assessment method, by setting up a causal inference model, distinguishes the real causes of risk from superficial correlations, avoids misjudgments and supports counterfactual stress testing, making risk decisions highly interpretable. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a three-dimensional structural diagram of the present invention. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0026] See also Figure 1 , the present invention provides a technical solution:

[0027] The financial risk assessment method based on big data includes the following steps:

[0028] S1: Obtain the structured data, unstructured data and real-time streaming data of the target entity through the multi-source heterogeneous data acquisition module;

[0029] S2: Construct an association relationship graph and model the risk propagation path of the target entity and its associated nodes based on a graph neural network;

[0030] S3: Through the multimodal data fusion module, feature alignment and joint representation learning are performed on structured data, unstructured text data, and time series data;

[0031] S4: Based on the causal inference model, separate the causal characteristics and confounding variables of the target entity risk event and generate causal risk factors;

[0032] S5: Input the graph risk propagation weight, multimodal feature vector and causal risk factor into the risk assessment model and output the comprehensive risk score of the target entity;

[0033] S6: Trigger early warnings or generate dynamic risk control strategies based on risk scores.

[0034] Through the dynamically constructed enterprise guarantee network and supply chain relationship map in the graph neural network, the risk transmission path between nodes can be quantified, hidden risk nodes can be effectively identified, and the chain reaction risk caused by the default of related enterprises can be reduced.

[0035] By integrating multi-source heterogeneous data such as financial statements, public opinion texts, and logistics trajectories, we can capture risk signals ignored by traditional models and improve the risk assessment coverage of small and micro enterprises and entities with weak credit records.

[0036] Through causal inference models, we can distinguish the real causes of risks from superficial correlations, avoid misjudgments, support counterfactual stress testing, and make risk decisions highly interpretable.

[0037] In this embodiment, the application of the graph neural network in S2 includes: constructing a dynamic graph through corporate equity relations, supply chain transaction records and guarantee networks, using a graph attention network to calculate the risk contagion weights between nodes, and identifying hidden risk propagation paths based on a random walk algorithm.

[0038] Through dynamic graphs, the risk transmission paths between related enterprises can be quantified, and chain reaction risks such as "broken guarantee chain" and "supply chain interruption" can be warned in advance. High-risk entities such as shadow banks and shell companies that are difficult to capture with traditional models can be identified, and hidden risk exposure can be reduced. Graph data can be updated in real time to capture equity changes or transaction anomalies, thereby improving the timeliness of risk assessment.

[0039] In this embodiment, the implementation of the multimodal data fusion module in S3 includes: extracting time series features from structured data, performing semantic vectorization on unstructured data using the BERT model, aligning heterogeneous features through a cross-modal Transformer model, and generating a joint embedding representation.

[0040] It integrates multimodal data such as financial statements (structured), public opinion news (unstructured), and logistics trajectories (time series), covers risk signals ignored by traditional models (such as the decline in debt repayment ability indicated by negative public opinion), supplements the credit assessment basis of small and micro enterprises through unstructured data (such as water and electricity bills, supply chain contracts), expands the service customer base, and eliminates data heterogeneity through cross-modal Transformer, improving the model's generalization ability for complex scenarios (such as cross-border payments).

[0041] In this embodiment, dual machine learning is used to eliminate confounding biases in observational data, counterfactual risk scenarios are constructed based on structural causal models, the causal effects of external shocks on target entities are quantified, and key causal chains are identified through causal discovery algorithms.

[0042] Distinguish between real risk triggers (such as policy adjustments) and superficial correlations (such as seasonal fluctuations), avoid excessive risk control or missed judgments caused by misjudgments, simulate the causal impact of extreme events (such as sudden interest rate increases and economic recessions) on asset portfolios, and optimize risk hedging strategies.

[0043] In this embodiment, the risk assessment model of S5 is an integrated model, including: taking the graph risk propagation weight as the input of the graph convolutional network, inputting the multimodal feature vector into the deep neural network branch, and making a joint decision with the causal risk factor and the above branch output through the weighted fusion layer.

[0044] By combining association networks, multimodal data, and causal factors, the accuracy of comprehensive risk assessment is improved (for example, the accuracy of high-risk customer identification is increased by 25%+), the bias of a single model is reduced through integrated learning, and it can adapt to complex financial scenarios.

[0045] In this embodiment, the data collection module in S1 further includes: obtaining pledge status data in real time through IoT devices, and performing cross-institutional joint modeling based on a federated learning framework to protect data privacy.

[0046] Through IoT data such as logistics GPS, warehouse temperature and humidity, the value fluctuations of collateral can be dynamically assessed (such as early warning of commodity price plunges). Federated learning enables cross-institutional data sharing modeling (such as inter-bank anti-money laundering collaboration) to avoid sensitive data leakage.

[0047] In this embodiment, the S5 also includes a risk explainability module: using the SHAP algorithm to perform attribution analysis on the risk assessment results, generating a key risk feature contribution report, and visually displaying the risk transmission path and causal chain through a knowledge graph.

[0048] Generate explainable risk reports (such as the contribution of key features), assist manual review and reduce compliance costs, and visualize risk transmission paths (such as illegal fund flows) through graphs to improve risk management efficiency.

[0049] In this embodiment, the dynamic risk control strategy of S6 includes: adjusting the credit limit or margin ratio in real time based on reinforcement learning, and generating a stress test simulation report for systemic risk scenarios.

[0050] Adjust credit limits or margin ratios in real time (e.g., automatically adjust limits based on changes in corporate cash flow), reduce default losses, generate stress testing reports (e.g., asset portfolio loss forecasts during an economic recession), and support management in formulating emergency strategies.

[0051] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0052] During operation, the multi-source heterogeneous data acquisition module is used to obtain the structured data, unstructured data and real-time streaming data of the target entity, and an association relationship graph is constructed. The risk propagation path of the target entity and its associated nodes is modeled based on the graph neural network. The multimodal data fusion module is used to perform feature alignment and joint representation learning on structured data, unstructured text data and time series data. Based on the causal inference model, the causal characteristics and confounding variables of the target entity risk events are separated to generate causal risk factors. The graph risk propagation weights, multimodal feature vectors and causal risk factors are input into the risk assessment model, and the comprehensive risk score of the target entity is output. According to the risk score, an early warning is triggered or a dynamic risk control strategy is generated.

[0053] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0054] 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 these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A financial risk assessment method based on big data, characterized by: The following steps are involved: S1: Obtain the structured data, unstructured data and real-time streaming data of the target entity through the multi-source heterogeneous data acquisition module; S2: Construct an association relationship graph and model the risk propagation path of the target entity and its associated nodes based on a graph neural network; S3: Through the multimodal data fusion module, feature alignment and joint representation learning are performed on structured data, unstructured text data, and time series data; S4: Based on the causal inference model, separate the causal characteristics and confounding variables of the target entity risk event and generate causal risk factors; S5: Input the graph risk propagation weight, multimodal feature vector and causal risk factor into the risk assessment model and output the comprehensive risk score of the target entity; S6: Trigger early warnings or generate dynamic risk control strategies based on risk scores.

2. The financial risk assessment method based on big data according to claim 1, characterized in that: The application of graph neural networks in S2 includes: building a dynamic graph through corporate equity relations, supply chain transaction records and guarantee networks, using graph attention networks to calculate risk contagion weights between nodes, and identifying hidden risk propagation paths based on random walk algorithms.

3. The financial risk assessment method based on big data according to claim 1, characterized in that: The implementation of the multimodal data fusion module in S3 includes: extracting temporal features from structured data, performing semantic vectorization on unstructured data using the BERT model, aligning heterogeneous features through a cross-modal Transformer model, and generating a joint embedding representation.

4. The financial risk assessment method based on big data according to claim 1, characterized in that: The causal inference model in S4 includes: using dual machine learning to eliminate confounding bias in observational data, constructing counterfactual risk scenarios based on structural causal models, quantifying the causal effects of external shocks on target entities, and identifying key causal chains through causal discovery algorithms.

5. The financial risk assessment method based on big data according to claim 1, characterized in that: The risk assessment model of S5 is an integrated model, which includes: taking the graph risk propagation weight as the input of the graph convolutional network, inputting the multimodal feature vector into the deep neural network branch, and making a joint decision with the causal risk factor and the above branch output through a weighted fusion layer.

6. The financial risk assessment method based on big data according to claim 1, characterized in that: The data collection module in S1 further includes: obtaining pledge status data in real time through IoT devices, and jointly modeling across institutions based on a federated learning framework to protect data privacy.

7. The financial risk assessment method based on big data according to claim 1, characterized in that: The S5 also includes a risk explainability module: using the SHAP algorithm to perform attribution analysis on risk assessment results, generating a key risk feature contribution report, and visually displaying the risk transmission path and causal chain through a knowledge graph.

8. The financial risk assessment method based on big data according to claim 1, characterized in that: The dynamic risk control strategy of S6 includes: adjusting credit limits or margin ratios in real time based on reinforcement learning, and generating stress test simulation reports for systemic risk scenarios.

Images