Risk assessment and early warning method for financial transaction

Abstract

The invention provides a risk assessment and early warning method for financial transactions, and the method comprises the steps: obtaining the historical market data, historical behavior data and historical flow data in the financial transactions, carrying out the preprocessing of the historical data, obtaining the target historical data, providing a data basis for the multi-dimensional risk assessment, and improving the risk assessment efficiency. Performing fusion processing on the target historical data, establishing a dynamic risk assessment rule according to a fusion processing result, ensuring timeliness and accuracy of risk assessment through establishment of the dynamic risk assessment rule, performing risk assessment on the current financial transaction data based on the dynamic risk assessment rule to obtain a risk assessment result, and performing risk assessment on the current financial transaction data based on the risk assessment result. On the basis of an automatic risk response mechanism, automatic risk response is performed on a risk assessment result, so that the risk can be handled in time, and the influence of risk events is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of financial transactions, and particularly relates to a method for risk assessment and early warning of financial transactions. Background Art

[0002] Financial risk assessment occupies an important position in the modern financial system. With the continuous development of the financial market, the complexity of risk management is also increasing continuously; through in-depth analysis of financial risk assessment, we can more comprehensively understand its importance. Risk assessment can not only help financial institutions identify potential problems, but also effectively reduce losses and improve operational efficiency; therefore, establishing a sound risk management system is crucial for ensuring the stable operation of institutions.

[0003] Along with the complexity of financial transactions and market fluctuations, the financial industry faces various risks, such as market risk, behavioral operation risk, liquidity risk, and credit risk. How to comprehensively evaluate multiple risks and ensure the timeliness and accuracy of risk assessment to reduce the impact of risk events is an urgent problem to be solved. Summary of the Invention

[0004] The present invention provides a method for risk assessment and early warning of financial transactions to solve the problems raised in the background art.

[0005] A method for risk assessment and early warning of financial transactions includes: S1: Obtain historical market data, historical behavior data, and historical liquidity data in financial transactions, and preprocess the historical data to obtain target historical data; S2: Perform fusion processing on the target historical data, and establish dynamic risk assessment rules according to the fusion processing results; S3: Perform risk assessment on the current financial transaction data based on the dynamic risk assessment rules to obtain a risk assessment result; S4: Perform automatic risk response to the risk assessment result based on an automated risk response mechanism.

[0006] Preferably, in the S1, obtaining historical market data, historical behavior data, and historical liquidity data in financial transactions, and preprocessing the historical data to obtain target historical data includes: Extract historical market transaction data from the publicly available information of financial transactions based on the key features of market data, extract historical behavior data from the publicly available information of financial transactions based on the key features of behavior data, and extract historical liquidity data from the publicly available information of financial transactions based on the key features of liquidity data; Perform data cleaning and data standardization on the historical market data, historical behavior data, and historical liquidity data respectively to obtain target historical data.

[0007] Preferably, in S2, the fusion processing of the target historical data includes: Based on the target historical data, a multi-modal data pipeline is established, and the data in the multi-modal pipeline is synchronized based on the Precision Time Protocol to obtain a synchronized data pipeline; Cross-pipeline data feature extraction is performed on the synchronized data pipeline according to time characteristics to obtain multiple synchronized multi-dimensional data feature groups; Mathematical transformation is performed on the synchronized multi-dimensional data feature groups based on financial indicators to obtain synchronized multi-dimensional composite index features; Based on financial sentiment information and combined with a comprehensive semantic analysis model, semantic analysis is performed on the synchronized multi-dimensional data features to obtain synchronized multi-dimensional semantic features; Obtain the historical result occurrence frequency in the historical financial risk assessment result, use the historical result occurrence frequency greater than the preset frequency as the historical target result, design multiple attention heads for the features based on the type of abnormal features corresponding to the temporary target result, and establish a weighted feature fusion mechanism based on the multiple attention heads for the features; Based on the weighted feature fusion mechanism, selectively weighted fusion is performed on the synchronized multi-dimensional data feature groups, synchronized multi-dimensional composite index features, and synchronized multi-dimensional semantic features to obtain initial fusion features, composite fusion features, and semantic fusion features; Based on the initial fusion features, composite fusion features, and semantic fusion features, determine the synchronized multi-dimensional fusion features, perform feature analysis on the synchronized multi-dimensional fusion features according to time characteristics, and obtain the correlation relationship between the synchronized multi-dimensional fusion features; Based on the correlation relationship, perform vertical fusion on the synchronized multi-dimensional fusion features to obtain the fusion processing result.

[0008] Preferably, cross-pipeline data feature extraction is performed on the synchronized data pipeline according to time characteristics to obtain multiple synchronized multi-dimensional data feature groups, including: Cross-pipeline data feature extraction is performed on the synchronized data pipeline according to time characteristics to obtain data features at the same time in different pipelines; Perform the same time marking on the data features at the same time to obtain a marked pipeline; Extract the data with the same time marking from the marked pipeline and combine the same time marking to obtain synchronized multi-dimensional data feature groups.

[0009] Preferably, in S2, the establishment of dynamic risk assessment rules according to the fusion processing result includes: Based on the multi-dimensional fusion feature groups obtained at the same time in the fusion processing result, determine the corresponding relationship between the fusion features and financial risks based on the multi-dimensional fusion feature groups, and establish a risk assessment example based on the corresponding relationship; Establish an initial static risk assessment rule based on a risk assessment example, and integrate and optimize the initial static risk assessment rule based on the association between risk assessment examples to obtain a target static risk assessment rule; Evolve the target static risk assessment rule for the multi-dimensional fusion feature group at all multiple times, and obtain an initial dynamic risk assessment rule based on the evolution process; Obtain the association situation between multi-dimensional fusion feature groups from the fusion processing result, and determine a first dynamic adjustment strategy based on the association situation; Obtain multi-dimensional dynamic features from the fusion processing result, and determine a second dynamic adjustment strategy based on the multi-dimensional dynamic features; Adjust the initial dynamic risk assessment rule based on the first dynamic adjustment strategy and the second dynamic adjustment strategy to obtain a final dynamic risk assessment rule.

[0010] Preferably, obtaining the association situation between multi-dimensional fusion feature groups from the fusion processing result and determining a first dynamic adjustment strategy based on the association situation includes: Obtain the dynamic evaluation threshold in the initial dynamic risk assessment rule; Obtain the association features related to the dynamic evaluation threshold from the association situation, and establish a mapping relationship between the association features and the dynamic evaluation threshold; Establish a first dynamic adjustment strategy for the initial dynamic risk assessment rule based on the mapping relationship.

[0011] Preferably, obtaining multi-dimensional dynamic features from the fusion processing result and determining a second dynamic adjustment strategy based on the multi-dimensional dynamic features includes: Obtain the dynamic weight of the evaluation index in the initial dynamic risk assessment rule; Obtain a dynamic feature group related to the evaluation index from the multi-dimensional dynamic features, and set a dynamic adjustment rule for the dynamic weight based on the fluctuation situation and occurrence frequency of the dynamic feature group; Determine a second dynamic adjustment strategy for the initial dynamic risk assessment rule based on the dynamic adjustment rule.

[0012] Preferably, in S3, performing a risk assessment on the current financial transaction data based on the dynamic risk assessment rule to obtain a risk assessment result, including: After preprocessing and multi-dimensional fusion of the current financial transaction data, obtain target current transaction data; Input the target current transaction data into the dynamic risk assessment rule to obtain a risk assessment result.

[0013] Preferably, the establishment of the automated risk response mechanism includes: Obtain abnormal financial features with risks, their corresponding countermeasures, and the effectiveness scores corresponding to the countermeasures from the publicly disclosed information of financial transactions; Perform relationship learning on the abnormal financial features, countermeasures, and effectiveness scores, and establish a feature-measurement comparison table that meets the preset effectiveness; Based on the feature-measurement comparison table, establish an automated risk response mechanism.

[0014] Preferably, in S4, based on the automated risk response mechanism, perform an automatic risk response to the risk assessment result, including: Input the risk assessment result into the automated risk response mechanism to obtain risk countermeasures; Perform an automated response to the risk countermeasures.

[0015] Compared with the prior art, the present invention has achieved the following beneficial effects: By obtaining historical market data, historical behavior data, and historical flow data in financial transactions, and preprocessing the historical data to obtain target historical data, providing a data basis for multi-dimensional risk assessment, performing fusion processing on the target historical data, establishing dynamic risk assessment rules based on the fusion processing results, ensuring the timeliness and accuracy of risk assessment through the establishment of dynamic risk assessment rules, performing risk assessment on the current financial transaction data based on the dynamic risk assessment rules to obtain a risk assessment result, and performing an automatic risk response to the risk assessment result based on the automated risk response mechanism, realizing timely response to risks and reducing the impact of risk events.

[0016] Other features and advantages of the present invention will be described in the subsequent specification, and part of them will become obvious from the specification or be understood by implementing the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the structure specifically pointed out in this application document.

[0017] The technical solution of the present invention will be further described in detail below through the accompanying drawings and embodiments. Description of the Drawings

[0018] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the accompanying drawings: Figure 1 It is a flowchart of a risk assessment and early warning method for a financial transaction in an embodiment of the present invention; Figure 2 It is a flowchart of obtaining a risk assessment result in an embodiment of the present invention; Figure 3 It is a flowchart of performing an automatic risk response in an embodiment of the present invention. Detailed Implementation Modes

[0019] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0020] Embodiment 1: The embodiment of the present invention provides a method for risk assessment and early warning of financial transactions, as Figure 1 shown, including: S1: Obtain historical market data, historical behavior data, and historical flow data in financial transactions, and preprocess the historical data to obtain target historical data; S2: Perform fusion processing on the target historical data, and establish a dynamic risk assessment rule according to the fusion processing result; S3: Perform risk assessment on the current financial transaction data based on the dynamic risk assessment rule to obtain a risk assessment result; S4: Based on an automated risk response mechanism, perform an automatic risk response to the risk assessment result.

[0021] In this embodiment, the historical market data includes order data, market data, etc., the historical behavior data includes buying and selling data, investment data, etc., and the historical flow data includes asset data and liability data.

[0022] In this embodiment, preprocessing the historical data means preprocessing the historical market data, historical behavior data, and historical flow data, and the preprocessing includes data cleaning, data standardization, etc.

[0023] In this embodiment, performing fusion processing on the target historical data realizes multi-dimensional fusion of the data.

[0024] In this embodiment, the automated risk response mechanism can perform automated early warning and corresponding automated operations according to the wind direction assessment result, such as freezing the account, blocking the transaction, etc.

[0025] The beneficial effects of the above design solution are as follows: By obtaining historical market data, historical behavior data, and historical flow data in financial transactions, and preprocessing the historical data to obtain target historical data, a data basis for multi-dimensional risk assessment is provided. Perform fusion processing on the target historical data, establish a dynamic risk assessment rule according to the fusion processing result, ensure the timeliness and accuracy of risk assessment by establishing the dynamic risk assessment rule, perform risk assessment on the current financial transaction data based on the dynamic risk assessment rule to obtain a risk assessment result, and based on an automated risk response mechanism, perform an automatic risk response to the risk assessment result to achieve timely response to risks and reduce the impact of risk events.

[0026] Embodiment 2: Based on Embodiment 1, an embodiment of the present invention provides a method for risk assessment and early warning of financial transactions. In S1, historical market data, historical behavior data, and historical flow data in financial transactions are obtained, and the historical data is preprocessed to obtain target historical data, including: Historical market transaction data is extracted from the publicly disclosed information of financial transactions based on the key features of market data, historical behavior data is extracted from the publicly disclosed information of financial transactions based on the key features of behavior data, and historical flow data is extracted from the publicly disclosed information of financial transactions based on the key features of flow data; Data cleaning and data standardization are respectively performed on the historical market data, historical behavior data, and historical flow data to obtain target historical data.

[0027] In this embodiment, the key features of market data are, for example, keywords of orders and market conditions; the key features of behavior data are, for example, keywords of buying and selling, investment, and the key features of flow data are, for example, asset replication keywords.

[0028] The beneficial effects of the above design scheme are: by obtaining historical market data, historical behavior data, and historical flow data in financial transactions, and respectively performing data cleaning and data standardization on the historical market data, historical behavior data, and historical flow data to obtain target historical data, a multi-dimensional data basis is provided for multi-dimensional risk assessment.

[0029] Embodiment 3: Based on Embodiment 1, an embodiment of the present invention provides a method for risk assessment and early warning of financial transactions. In S2, fusion processing is performed on the target historical data, including: Based on the target historical data, a multi-modal data pipeline is established, and data synchronization is performed on the multi-modal pipeline data based on the Precision Time Protocol to obtain a synchronized data pipeline; Cross-pipeline data feature extraction is performed on the synchronized data pipeline according to time characteristics to obtain multiple synchronized multi-dimensional data feature groups; Mathematical transformation is performed on the synchronized multi-dimensional data feature groups based on financial indicators to obtain synchronized multi-dimensional composite index features; Based on financial sentiment information, combined with a comprehensive semantic analysis model, semantic analysis is performed on the synchronized multi-dimensional data features to obtain synchronized multi-dimensional semantic features; The historical result occurrence frequency in the historical financial risk assessment results is obtained, and those with a historical result occurrence frequency greater than the preset frequency are used as historical target results. Based on the types of abnormal features corresponding to the temporary target results, multiple attention heads for features are designed, and a weighted feature fusion mechanism is established based on the multiple attention heads for features; Based on the weighted feature fusion mechanism, selectively weighted fusion is performed on the synchronous multi-dimensional data feature group, synchronous multi-dimensional composite index features, and synchronous multi-dimensional semantic features to obtain initial fusion features, composite fusion features, and semantic fusion features; Based on the initial fusion features, composite fusion features, and semantic fusion features, determine synchronous multi-dimensional fusion features, perform feature analysis on the synchronous multi-dimensional fusion features according to time features, and obtain the correlation relationship between the synchronous multi-dimensional fusion features; Based on the correlation relationship, perform vertical fusion on the synchronous multi-dimensional fusion features to obtain a fusion processing result.

[0030] In this embodiment, the multi-modal data pipeline includes market, behavior, and flow pipelines, and more detailed pipelines derived therefrom.

[0031] In this embodiment, the synchronous data pipeline realizes the synchronization of data between pipelines. For example, trading behaviors are carried out in the market at the same time.

[0032] In this embodiment, financial indicators include, for example, volatility, imbalance rate, behavior complexity, market preference degree, etc.

[0033] In this embodiment, synchronous multi-dimensional semantic features can represent the preferences, emotions, etc. of the current market.

[0034] In this embodiment, the synchronous multi-dimensional data feature group, synchronous multi-dimensional composite index features, and synchronous multi-dimensional semantic features respectively represent financial information from three aspects: basic, composite index, and semantics.

[0035] In this embodiment, the correlation relationship between synchronous multi-dimensional fusion features includes the correlation between the front-back order of different time periods.

[0036] In this embodiment, the fusion processing result realizes the all-round fusion of historical target features.

[0037] The beneficial effects of the above design solution are as follows: By obtaining a synchronous data pipeline based on the target historical data, multiple synchronous multi-dimensional data feature groups are obtained, realizing information synchronization of data in different dimensions and providing a synchronization basis for further data fusion. By extracting cross-pipeline data features from the synchronous data pipeline according to time characteristics, multiple synchronous multi-dimensional data feature groups are obtained. Based on financial indicators, mathematical transformations are performed on the synchronous multi-dimensional data feature groups to obtain synchronous multi-dimensional composite index features. Based on financial sentiment information and combined with a comprehensive semantic analysis model, semantic analysis is performed on the synchronous multi-dimensional data features to obtain synchronous multi-dimensional semantic features, realizing the horizontal derivation of features and obtaining the historical result occurrence frequency in the historical financial risk assessment result. The historical results with an occurrence frequency greater than the preset frequency are used as historical target results. Based on the types of abnormal features corresponding to the temporary target results, multiple attention heads for features are designed. Based on the multiple attention heads for features, a weighted feature fusion mechanism is established. Based on the weighted feature fusion mechanism, selective weighted fusion is performed on the synchronous multi-dimensional data feature groups, synchronous multi-dimensional composite index features, and synchronous multi-dimensional semantic features respectively to obtain initial fusion features, composite fusion features, and semantic fusion features, realizing the weighted sum and extraction of intermediate features. Based on the initial fusion features, composite fusion features, and semantic fusion features, synchronous multi-dimensional fusion features are determined. Feature analysis is performed on the synchronous multi-dimensional fusion features according to time characteristics to obtain the correlation relationship between the synchronous multi-dimensional fusion features. Based on the correlation relationship, vertical fusion is performed on the synchronous multi-dimensional fusion features to obtain a fusion processing result, realizing the vertical fusion of features. Finally, the all-round analysis and fusion of historical target information are realized, ensuring that the obtained fusion processing result can more comprehensively and accurately represent the characteristics of financial data.

[0038] Embodiment 4: Based on Embodiment 3, an embodiment of the present invention provides a risk assessment and early warning method for financial transactions. Cross-pipeline data feature extraction is performed on the synchronous data pipeline according to time characteristics to obtain multiple synchronous multi-dimensional data feature groups, including: Cross-pipeline data feature extraction is performed on the synchronous data pipeline according to time characteristics to obtain data features at the same time in different pipelines; The same time mark is given to the data features at the same time to obtain a marked pipeline; Extract the data with the same time mark from the marked pipeline and combine the same time mark to obtain a synchronous multi-dimensional data feature group.

[0039] The beneficial effects of the above design solution are as follows: By extracting data after further marking the data in the pipeline and combining the same time mark to obtain a synchronous multi-dimensional data feature group, the time consistency of the obtained synchronous multi-dimensional data feature group is ensured, providing a basis for data fusion.

[0040] Example 5: Based on Example 1, an embodiment of the present invention provides a risk assessment and early warning method for financial transactions. In S2, establishing a dynamic risk assessment rule according to the fusion processing result includes: Based on the multi-dimensional fusion feature group obtained at the same time in the fusion processing result, determining the correspondence between the fusion feature and the financial risk based on the multi-dimensional fusion feature group, and establishing a risk assessment example based on the correspondence; Establishing an initial static risk assessment rule based on the risk assessment example, and integrating and optimizing the initial static risk assessment rule based on the association between risk assessment examples to obtain a target static risk assessment rule; Evolving the target static risk assessment rule of the multi-dimensional fusion feature group at all multiple times, and obtaining an initial dynamic risk assessment rule based on the evolution process; Obtaining the association situation between multi-dimensional fusion feature groups from the fusion processing result, and determining a first dynamic adjustment strategy based on the association situation; Obtaining multi-dimensional dynamic features from the fusion processing result, and determining a second dynamic adjustment strategy based on the multi-dimensional dynamic features; Adjusting the initial dynamic risk assessment rule based on the first dynamic adjustment strategy and the second dynamic adjustment strategy to obtain a final dynamic risk assessment rule.

[0041] In this embodiment, at least one risk assessment example corresponds to one time point.

[0042] In this embodiment, integrating and optimizing the initial static risk assessment rule is, for example, obtaining the target static risk assessment rule by taking the average value, superimposing, etc.

[0043] In this embodiment, the evolution process realizes the coherence of the rules to obtain an initial dynamic risk assessment rule.

[0044] In this embodiment, based on the first dynamic adjustment strategy and the second dynamic adjustment strategy, adjusting the initial dynamic risk assessment rule by adding a dynamic adjustment strategy realizes the purpose of real-time adjustment of the dynamic rule according to specific feedback.

[0045] The beneficial effects of the above design solution are as follows: By obtaining the multi-dimensional fusion feature group at the same time from the fusion processing result, determining the corresponding relationship between the fusion feature and the financial risk based on the multi-dimensional fusion feature group, establishing a risk assessment example based on the corresponding relationship, establishing an initial static risk assessment rule based on the risk assessment example, and integrating and optimizing the initial static risk assessment rule based on the association between risk assessment examples to obtain the target static risk assessment rule, ensuring the accuracy of the obtained Alibaba static rule. By evolving the target static risk assessment rule of the multi-dimensional fusion feature group at all multiple times, an initial dynamic risk assessment rule is obtained based on the evolution process, realizing the establishment of a dynamic rule. The association situation between multi-dimensional fusion feature groups is obtained from the fusion processing result, and the first dynamic adjustment strategy is determined based on the association situation; multi-dimensional dynamic features are obtained from the fusion processing result, and the second dynamic adjustment strategy is determined based on the multi-dimensional dynamic features; based on the first dynamic adjustment strategy and the second dynamic adjustment strategy, the initial dynamic risk assessment rule is adjusted to obtain the final dynamic risk assessment rule, ensuring the timeliness and accuracy of risk assessment through the establishment of the dynamic risk assessment rule.

[0046] Embodiment 6: Based on Embodiment 5, an embodiment of the present invention provides a method for risk assessment and early warning of financial transactions. Obtaining the association situation between multi-dimensional fusion feature groups from the fusion processing result, and determining the first dynamic adjustment strategy based on the association situation, including: Obtain the dynamic assessment threshold in the initial dynamic risk assessment rule; Obtain the associated features related to the dynamic assessment threshold from the association situation, and establish a mapping relationship between the associated features and the dynamic assessment threshold; Based on the mapping relationship, establish the first dynamic adjustment strategy for the initial dynamic risk assessment rule.

[0047] In this embodiment, the dynamic assessment threshold can be adjusted in real time, for example, the limit on the transaction amount.

[0048] The beneficial effects of the above design solution are as follows: By obtaining the dynamic assessment threshold in the initial dynamic risk assessment rule, obtaining the associated features related to the dynamic assessment threshold from the association situation, establishing a mapping relationship between the associated features and the dynamic assessment threshold, and establishing the first dynamic adjustment strategy for the initial dynamic risk assessment rule based on the mapping relationship, a real-time adjustment strategy for the dynamic threshold is established, achieving the purpose of real-time adjustment of the dynamic rule according to specific feedback.

[0049] Embodiment 7: Based on Embodiment 5, an embodiment of the present invention provides a risk assessment and early warning method for financial transactions, obtaining multi-dimensional dynamic features from the fusion processing result, and determining a second dynamic adjustment strategy based on the multi-dimensional dynamic features, including: Obtain the dynamic weights of the evaluation indicators in the initial dynamic risk assessment rule; Obtain a dynamic feature group related to the evaluation indicator from the multi-dimensional dynamic features, and set a dynamic adjustment rule for the dynamic weight based on the fluctuation situation and occurrence frequency of the dynamic feature group; Determine a second dynamic adjustment strategy for the initial dynamic risk assessment rule based on the dynamic adjustment rule.

[0050] In this embodiment, the dynamic weight characterizes the importance of each evaluation indicator for risk assessment.

[0051] The beneficial effect of the above design solution is: by obtaining the dynamic weights of the evaluation indicators in the initial dynamic risk assessment rule, obtaining a dynamic feature group related to the evaluation indicator from the multi-dimensional dynamic features, setting a dynamic adjustment rule for the dynamic weight based on the fluctuation situation and occurrence frequency of the dynamic feature group, and determining a second dynamic adjustment strategy for the initial dynamic risk assessment rule based on the dynamic adjustment rule, the purpose of real-time adjustment of the dynamic rule according to specific feedback is achieved.

[0052] Embodiment 8: Based on Embodiment 1, an embodiment of the present invention provides a risk assessment and early warning method for financial transactions. As Figure 2 shown, in S3, risk assessment is performed on the current financial transaction data based on the dynamic risk assessment rule, and a risk assessment result is obtained, including: After preprocessing and multi-dimensional fusion of the current financial transaction data, target current transaction data is obtained; Input the target current transaction data into the dynamic risk assessment rule to obtain a risk assessment result.

[0053] In this embodiment, the risk assessment result includes the score of the risk assessment, the risk type, and the specific data situation of the risk.

[0054] The beneficial effect of the above design solution is: by preprocessing and multi-dimensional fusion of the current financial transaction data to obtain target current transaction data, and inputting the target current transaction data into the dynamic risk assessment rule to obtain a risk assessment result, the accuracy and timeliness of the risk assessment result are ensured by the dynamic risk assessment rule.

[0055] Embodiment 9: Based on Embodiment 1, an embodiment of the present invention provides a risk assessment and early warning method for financial transactions. The establishment of the automated risk response mechanism includes: Obtain the abnormal financial features at risk and their corresponding countermeasures, as well as the effectiveness scores corresponding to the countermeasures, from the publicly disclosed information of financial transactions; Conduct relationship learning on the abnormal financial features, countermeasures, and effectiveness scores to establish a feature-measure correspondence table that meets the preset effectiveness; Establish an automated risk response mechanism based on the feature-measure correspondence table.

[0056] The beneficial effects of the above design scheme are as follows: By obtaining the abnormal financial features at risk and their corresponding countermeasures, as well as the effectiveness scores corresponding to the countermeasures, from the publicly disclosed information of financial transactions, conducting relationship learning on the abnormal financial features, countermeasures, and effectiveness scores, establishing a feature-measure correspondence table that meets the preset effectiveness, and establishing an automated risk response mechanism based on the feature-measure correspondence table, through the analysis and selection of historical financial information, obtain countermeasures that meet the requirements, and apply these countermeasures to the automated risk response mechanism to provide a response mechanism for the subsequent response to financial risks.

[0057] Example 10: Based on Example 1, an embodiment of the present invention provides a method for risk assessment and early warning of financial transactions, as Figure 3 shown. In S4, based on the automated risk response mechanism, perform an automatic risk response to the risk assessment result, including: Input the risk assessment result into the automated risk response mechanism to obtain risk countermeasures; Perform an automated response to the risk countermeasures.

[0058] In this embodiment, when the risk assessment result is that no risk occurs, the corresponding risk countermeasure is to do nothing.

[0059] The beneficial effects of the above design scheme are as follows: By inputting the risk assessment result into the automated risk response mechanism to obtain risk countermeasures and performing an automated response to the risk countermeasures, it realizes the timely response to risks and reduces the impact of risk events.

[0060] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of this application document and its equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A risk assessment and early warning method for financial transactions, characterized in that: include: S1: Obtain historical market data, historical behavior data, and historical flow data in financial transactions, and pre-process the historical data to obtain target historical data; S2: performing fusion processing on the target historical data, and establishing dynamic risk assessment rules according to the fusion processing results; S3: Perform risk assessment on current financial transaction data based on dynamic risk assessment rules to obtain risk assessment results; S4: Based on the automated risk response mechanism, automatic risk response is performed on the risk assessment results.

2. A risk assessment and early warning method for financial transactions according to claim 1, characterized in that: In S1, historical market data, historical behavior data and historical flow data in financial transactions are obtained, and the historical data are preprocessed to obtain target historical data, including: Based on the key features of market data, historical market transaction data is extracted from the public information of financial transactions; based on the key features of behavior data, historical behavior data is extracted from the public information of financial transactions; based on the key features of flow data, historical flow data is extracted from the public information of financial transactions; Data cleaning and data standardization are performed on historical market data, historical behavior data and historical flow data respectively to obtain target historical data.

3. A risk assessment and early warning method for financial transactions according to claim 1, characterized in that: In S2, the target historical data is subjected to fusion processing, including: Based on the target historical data, a multimodal data pipeline is established, and data synchronization is performed on the multimodal pipeline data based on a precise time protocol to obtain a synchronized data pipeline; Extracting cross-pipeline data features from the synchronous data pipeline according to time features to obtain multiple synchronous multi-dimensional data feature groups; Performing mathematical transformation on the synchronous multi-dimensional data feature group based on financial indicators to obtain synchronous multi-dimensional composite indicator features; Based on financial sentiment information and combined with a comprehensive semantic analysis model, semantic analysis is performed on synchronous multi-dimensional data features to obtain synchronous multi-dimensional semantic features; Obtain the frequency of occurrence of historical results in historical financial risk assessment results, take the historical results with a frequency greater than a preset frequency as historical target results, design multiple attention heads for features based on the type of abnormal features corresponding to the temporary target results, and establish a weighted feature fusion mechanism based on the multiple attention heads for features; Based on the weighted feature fusion mechanism, the synchronous multi-dimensional data feature group, the synchronous multi-dimensional composite indicator feature and the synchronous multi-dimensional semantic feature are selectively weighted and fused to obtain the initial fusion feature, the composite fusion feature and the semantic fusion feature; Determine the synchronous multi-dimensional fusion features based on the initial fusion features, the composite fusion features and the semantic fusion features, perform feature analysis on the synchronous multi-dimensional fusion features according to the time features, and obtain the correlation between the synchronous multi-dimensional fusion features; Based on the association relationship, the synchronous multi-dimensional fusion features are vertically fused to obtain a fusion processing result.

4. A risk assessment and early warning method for financial transactions according to claim 3, characterized in that: The cross-pipeline data feature extraction is performed on the synchronous data pipeline according to the time feature to obtain multiple synchronous multi-dimensional data feature groups, including: Extracting cross-pipeline data features from the synchronous data pipeline according to time features to obtain data features of different pipelines at the same time; The data features at the same time are marked with the same time to obtain a marking pipeline; Data with the same time tag in the tagging pipeline are extracted, and the same time tag is combined to obtain a synchronized multi-dimensional data feature group.

5. A risk assessment and early warning method for financial transactions according to claim 1, characterized in that: In S2, dynamic risk assessment rules are established according to the fusion processing results, including: Based on the multi-dimensional fusion feature group obtained at the same time from the fusion processing result, the corresponding relationship between the fusion feature and the financial risk is determined based on the multi-dimensional fusion feature group, and based on the corresponding relationship, a risk assessment example is established; Establishing initial static risk assessment rules based on risk assessment examples, and integrating and optimizing the initial static risk assessment rules based on associations between risk assessment examples to obtain target static risk assessment rules; Evolving the target static risk assessment rules of the multi-dimensional fusion feature group at all multiple times, and obtaining the initial dynamic risk assessment rules based on the evolution process; Obtaining correlation between the multi-dimensional fusion feature groups from the fusion processing result, and determining a first dynamic adjustment strategy based on the correlation; Obtaining a multi-dimensional dynamic feature from the fusion processing result, and determining a second dynamic adjustment strategy based on the multi-dimensional dynamic feature; Based on the first dynamic adjustment strategy and the second dynamic adjustment strategy, the initial dynamic risk assessment rule is adjusted to obtain a final dynamic risk assessment rule.

6. A risk assessment and early warning method for financial transactions according to claim 5, characterized in that: Obtaining correlations between the multi-dimensional fusion feature groups from the fusion processing result, and determining a first dynamic adjustment strategy based on the correlations, including: Obtaining a dynamic assessment threshold in the initial dynamic risk assessment rule; Acquire correlation features related to the dynamic evaluation threshold from the correlation situation, and establish a mapping relationship between the correlation features and the dynamic evaluation threshold; A first dynamic adjustment strategy for the initial dynamic risk assessment rule is established based on the mapping relationship.

7. A risk assessment and early warning method for financial transactions according to claim 5, characterized in that: Obtaining a multi-dimensional dynamic feature from the fusion processing result, and determining a second dynamic adjustment strategy based on the multi-dimensional dynamic feature, including: Obtaining the dynamic weights of the evaluation indicators in the initial dynamic risk evaluation rules; Acquire a dynamic feature group related to the evaluation index from the multi-dimensional dynamic features, and set a dynamic adjustment rule for the dynamic weight based on the fluctuation and occurrence frequency of the dynamic feature group; A second dynamic adjustment strategy for the initial dynamic risk assessment rule is determined based on the dynamic adjustment rule.

8. A risk assessment and early warning method for financial transactions according to claim 1, characterized in that: In S3, risk assessment is performed on the current financial transaction data based on the dynamic risk assessment rules to obtain risk assessment results, including: After preprocessing and multi-dimensional fusion of current financial transaction data, the target current transaction data is obtained; The target current transaction data is input into the dynamic risk assessment rules to obtain the risk assessment result.

9. A risk assessment and early warning method for financial transactions according to claim 1, characterized in that: The establishment of the automated risk response mechanism includes: Obtain risky abnormal financial features and their corresponding countermeasures from public information on financial transactions, as well as the effectiveness scores of the countermeasures; Conduct relationship learning on abnormal financial features, countermeasures and effect scores, and establish a feature-measure comparison table that meets the preset effects; An automated risk response mechanism is established based on the feature-measure comparison table.

10. A financial transaction risk assessment and early warning method according to claim 1, characterized in that: In S4, based on the automated risk response mechanism, an automated risk response is performed on the risk assessment results, including: Input the risk assessment results into the automated risk response mechanism to obtain risk response measures; Automated responses to the risk responses.

Images