Financial risk early warning system and method and storage medium

Abstract

The invention relates to the technical field of financial risk analysis, and particularly discloses a financial risk early warning system and method and a storage medium, and the method comprises the steps: firstly obtaining a newest financial supervision file and newest financial punishment information, and extracting a risk index set and a punishment risk point set from the newest financial supervision file and the newest financial punishment information; and then semantic embedding coding is performed on the risk indicator set and the punishment risk point set by using a natural language processing technology based on deep learning, a financial risk assessment point semantic embedding representation set is constructed based on an association relationship between the risk indicators and punishment risk points, and then a financial risk assessment point semantic embedding representation set is constructed. Through rapid semantic query interaction between the financial service and each financial risk assessment point, potential risk information of the financial service is mined, so that intelligent identification and early warning of financial service risks are realized. According to the invention, the intelligent level of financial risk early warning can be effectively improved, the capability of identifying financial business risks is enhanced, and the compliance and security of financial businesses are ensured.

Description

Technical Field

[0001] The present application relates to the technical field of financial risk analysis, and more specifically, to a financial risk early warning system, method and storage medium. Background Art

[0002] With the rapid development of the financial industry and the continuous updating of laws and regulations, financial institutions are facing increasing compliance and risk control challenges. In order to ensure the security and compliance of financial business, it is essential to identify and warn potential financial risks in a timely and accurate manner. Traditional financial risk management methods usually rely on manual interpretation of the latest financial regulatory documents and penalty information. This method has a lengthy and inefficient internal information transmission process, which is not only inefficient, but also easily affected by personal experience and background, leading to deviations in the understanding of policies, which may lead to unnecessary administrative and economic penalties, and even cause systemic financial risks.

[0003] In this regard, the invention patent with publication number CN116611917A proposes a financial risk early warning method. The method obtains the latest financial regulatory documents and the latest financial penalty information, and parses them using text processing technology, thereby providing risk warnings for financial services based on the parsed risk indicator set and penalty risk point set, thereby improving the security of financial services.

[0004] However, in the existing technology, it mainly relies on traditional keyword rule matching and word frequency statistical analysis methods to process text data, which may not be flexible and accurate enough when facing the complex and changing financial market. For example, due to the lack of understanding of deep semantic information and contextual relationships in the text, this traditional text processing method has limited processing capabilities for deep semantic associations between risk indicators and penalty risk points, the diversity, complexity and ambiguity of financial business language expressions, etc., which may affect the accuracy of financial risk assessment and the timeliness of early warning.

[0005] Therefore, an optimized financial risk early warning system, method and storage medium are needed to solve the above technical problems. Summary of the invention

[0006] In order to solve the above technical problems, according to one aspect of the present application, a financial risk early warning method is provided, which includes: obtaining the latest financial regulatory documents and the latest financial penalty information; parsing the latest financial regulatory documents to obtain a risk indicator set; parsing the latest financial penalty information to obtain a penalty risk point set; based on the risk indicator set and the penalty risk point set, risk early warning for financial business is performed; Based on the risk indicator set and the penalty risk point set, a risk warning is provided for the financial business, including: Semantically embedding and coding each risk indicator in the risk indicator set and each penalty risk point in the penalty risk point set to obtain a set of risk indicator semantic embedding coding vectors and a set of penalty risk point semantic embedding coding vectors; Merging the set of risk indicator semantic embedding coding vectors and the set of penalty risk point semantic embedding coding vectors to obtain a set of financial risk assessment point semantic embedding coding vectors; Performing semantic coding on the financial service to obtain a financial service semantic coding vector; Performing fast matching-oriented semantic query encoding on the set of the financial business semantic encoding vector and the financial risk assessment point semantic embedding encoding vector to obtain a financial business-financial risk assessment point query response semantic encoding vector; Based on the financial business-financial risk assessment point query response semantic coding vector, a financial risk analysis result is determined.

[0007] Specifically, the set of risk indicator semantic embedding coding vectors and the set of penalty risk point semantic embedding coding vectors are combined to obtain a set of financial risk assessment point semantic embedding coding vectors, including: Calculating the cosine similarity between each risk indicator semantic embedding coding vector in the set of risk indicator semantic embedding coding vectors and each penalty risk point semantic embedding coding vector in the set of penalty risk point semantic embedding coding vectors to obtain a risk indicator-penalty risk point semantic association topological matrix; The risk indicator semantic embedding coding vectors and the penalty risk point semantic embedding coding vectors corresponding to the cosine similarity greater than a preset threshold in the risk indicator-penalty risk point semantic association topological matrix are cascaded to obtain a set of financial risk assessment point semantic embedding coding vectors.

[0008] Specifically, semantic encoding is performed on the financial service to obtain a financial service semantic encoding vector, including: The financial service is semantically encoded using a financial service semantic encoder based on the Bert model to obtain the financial service semantic encoding vector.

[0009] Specifically, the set of the financial business semantic coding vector and the financial risk assessment point semantic embedding coding vector is subjected to fast matching-oriented semantic query coding to obtain a financial business-financial risk assessment point query response semantic coding vector, including: Determine the positioning center financial risk assessment point semantic embedding coding vector of the set of financial risk assessment point semantic embedding coding vectors based on the semantic difference between the financial business semantic encoding vector and each financial risk assessment point semantic embedding coding vector in the set of financial risk assessment point semantic embedding coding vectors; Determining a fine-grained matching search window for financial risk assessment points based on the semantic embedding coding vector of the positioning center financial risk assessment point; Fine-grained semantic query response interactive encoding is performed on the financial business semantic encoding vector and each financial risk assessment point semantic embedding encoding vector in the financial risk assessment point fine-grained matching search window to obtain the financial business-financial risk assessment point query response semantic encoding vector.

[0010] Specifically, based on the semantic difference between the financial business semantic coding vector and each financial risk assessment point semantic embedding coding vector in the set of financial risk assessment point semantic embedding coding vectors, determining the positioning center financial risk assessment point semantic embedding coding vector of the set of financial risk assessment point semantic embedding coding vectors includes: Calculating the cross entropy of the financial business semantic encoding vector relative to each financial risk assessment point semantic embedding encoding vector in the set of financial risk assessment point semantic embedding encoding vectors to obtain a set of financial risk assessment point fast query positioning factors; The financial risk assessment point semantic embedding coding vector corresponding to the minimum value in the set of the financial risk assessment point quick query positioning factors is used as the positioning center financial risk assessment point semantic embedding coding vector.

[0011] Specifically, the vector of the center position of the financial risk assessment point fine-grained matching search window is the semantic embedding coding vector of the positioned center financial risk assessment point, and the semantic embedding coding vectors of each financial risk assessment point in the financial risk assessment point fine-grained matching search window are defined as fine-grained query financial risk assessment point semantic embedding coding vectors.

[0012] Specifically, fine-grained semantic query response interactive coding is performed on the financial business semantic coding vector and each financial risk assessment point semantic embedding coding vector in the financial risk assessment point fine-grained matching search window to obtain the financial business-financial risk assessment point query response semantic coding vector, including: The financial business semantic coding vector is linearly transformed to obtain a query vector and a value vector, and each fine-grained query financial risk assessment point semantic embedding coding vector in the financial risk assessment point fine-grained matching search window is used as a key vector, and the query vector, the value vector and the key vector are input into a fine-grained query coding module based on a heterogeneous converter structure to obtain the financial business-financial risk assessment point query response semantic coding vector.

[0013] Specifically, based on the financial business-financial risk assessment point query response semantic coding vector, determining the financial risk analysis result includes: The financial business-financial risk assessment point query response semantic encoding vector is input into a classifier-based financial risk analysis module to obtain the financial risk analysis result, and the financial risk analysis result is used to indicate whether there is a financial risk in the financial business.

[0014] According to another aspect of the present application, a financial risk early warning system is provided, which includes: A risk data semantic embedding coding module is used to perform semantic embedding coding on each risk indicator in the risk indicator set and each penalty risk point in the penalty risk point set to obtain a set of risk indicator semantic embedding coding vectors and a set of penalty risk point semantic embedding coding vectors; A risk data association merging module, used for merging the set of risk indicator semantic embedding coding vectors and the set of penalty risk point semantic embedding coding vectors to obtain a set of financial risk assessment point semantic embedding coding vectors; A financial service semantic coding module, used for semantically coding financial services to obtain financial service semantic coding vectors; A fast matching semantic query encoding module, used for performing fast matching-oriented semantic query encoding on the set of the financial business semantic encoding vector and the financial risk assessment point semantic embedding encoding vector to obtain a financial business-financial risk assessment point query response semantic encoding vector; The analysis result generating module is used to determine the financial risk analysis result based on the financial business-financial risk assessment point query response semantic coding vector.

[0015] According to another aspect of the present application, a storage medium is provided, on which a financial risk warning program is stored. When the financial risk warning program is executed by a processor, the financial risk warning method as described above is implemented.

[0016] This application has at least the following technical effects: Compared with the prior art, the financial risk early warning system, method and storage medium provided by the present application first obtain the latest financial regulatory documents and the latest financial penalty information, and extract the risk indicator set and penalty risk point set from them, and then use the natural language processing technology based on deep learning to semantically embed the risk indicator set and the penalty risk point set, and based on the correlation between the risk indicator and the penalty risk point, construct a set of semantic embedding representations of financial risk assessment points, and then through rapid semantic query interaction between financial business and each financial risk assessment point, to dig out the potential risk information of financial business, so as to realize intelligent identification and early warning of financial business risks. This application can effectively improve the intelligence level of financial risk early warning, enhance the ability to identify financial business risks, and ensure the compliance and security of financial business. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 Flow chart of a financial risk warning method according to an embodiment of the present application.

[0019] Figure 2 Schematic diagram of data flow of the financial risk warning method according to an embodiment of the present application.

[0020] Figure 3 This is a flowchart of step S120 in the financial risk warning method according to an embodiment of the present application.

[0021] Figure 4 This is a flowchart of step S140 in the financial risk warning method according to an embodiment of the present application.

[0022] Figure 5 It is a block diagram of a financial risk early warning system according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0024] It should be noted that the acquisition and processing of all information or data in this application are carried out in compliance with the relevant national data protection laws and policies and with the authorization given by the owner of the corresponding device.

[0025] In response to the problems existing in the background technology, this application proposes an optimized financial risk warning method, which first obtains the latest financial regulatory documents and the latest financial penalty information, and extracts risk indicator sets and penalty risk point sets from them, and then uses natural language processing technology based on deep learning to semantically embed and encode risk indicator sets and penalty risk point sets, and based on the correlation between risk indicators and penalty risk points, constructs a set of semantic embedding representations of financial risk assessment points, and then conducts rapid semantic query interaction on financial business and each financial risk assessment point to dig out potential risk information of financial business, thereby realizing intelligent identification and warning of financial business risks. In this way, the intelligence level of financial risk warning can be effectively improved, the ability to identify financial business risks can be enhanced, and the compliance and security of financial business can be ensured.

[0026] Specifically, Figure 1 Flow chart of a financial risk warning method according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the financial risk early warning method according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the financial risk warning method according to the embodiment of the present application includes the steps of: S110, respectively performing semantic embedding coding on each risk indicator in the risk indicator set and each penalty risk point in the penalty risk point set to obtain a set of risk indicator semantic embedding coding vectors and a set of penalty risk point semantic embedding coding vectors; S120, merging the set of risk indicator semantic embedding coding vectors and the set of penalty risk point semantic embedding coding vectors to obtain a set of financial risk assessment point semantic embedding coding vectors; S130, performing semantic coding on the financial business to obtain a financial business semantic coding vector; S140, performing fast matching-oriented semantic query coding on the set of financial business semantic coding vectors and the set of financial risk assessment point semantic embedding coding vectors to obtain a financial business-financial risk assessment point query response semantic coding vector; S150, determining the financial risk analysis result based on the financial business-financial risk assessment point query response semantic coding vector.

[0027] In the above-mentioned financial risk early warning method, the step S110 performs semantic embedding coding on each risk indicator in the risk indicator set and each penalty risk point in the penalty risk point set to obtain a set of risk indicator semantic embedding coding vectors and a set of penalty risk point semantic embedding coding vectors. Among them, considering that the traditional keyword matching and word frequency statistical analysis methods are difficult to capture the implicit semantics of the text when processing text data, it may lead to misunderstandings of the complex associations between risk indicators and penalty risk points. Therefore, in order to more accurately process and understand financial risks, this application uses natural language processing technology based on deep learning to perform semantic embedding coding on each risk indicator in the risk indicator set and each penalty risk point in the penalty risk point set, so as to map each risk indicator and penalty risk point to a high-dimensional semantic feature space, and convert it into a vector representation with rich semantic information, so as to facilitate the semantic association analysis between risk indicators and penalty risk points in the high-dimensional semantic feature space, and obtain a deeper semantic understanding of risk indicators and penalty risk points. In an embodiment of the present application, the Word2Vec model is used to perform semantic embedding coding processing on each risk indicator and each penalty risk point, respectively, to obtain a set of risk indicator semantic embedding coding vectors and a set of penalty risk point semantic embedding coding vectors. The Word2Vec model is a widely used word embedding model that can map words into a continuous vector space through training, so that semantically similar words are close to each other in the vector space, thereby effectively capturing the potential connection between risk indicators and penalty risk points, and providing more accurate data support for subsequent risk assessment and early warning.

[0028] In the above-mentioned financial risk early warning method, the step S120 combines the set of risk indicator semantic embedded coding vectors and the set of penalty risk point semantic embedded coding vectors to obtain a set of financial risk assessment point semantic embedded coding vectors. Among them, risk indicators usually reflect potential risk areas in the daily operations of financial institutions, while penalty risk points are directly related to violations or regulatory actions that have occurred. In the technical solution of the present application, by merging the set of risk indicator semantic embedded coding vectors and the set of penalty risk point semantic embedded coding vectors, a more comprehensive and integrated risk view can be provided, covering everything from potential risks to actual risk events.

[0029] Figure 3 FIG. 1 is a flow chart of step S120 in the financial risk early warning method according to an embodiment of the present application. Figure 3As shown, the step S120 includes: S121, calculating the cosine similarity between each risk indicator semantic embedding coding vector in the set of risk indicator semantic embedding coding vectors and each penalty risk point semantic embedding coding vector in the set of penalty risk point semantic embedding coding vectors to obtain a risk indicator-penalty risk point semantic association topological matrix; S122, cascading the risk indicator semantic embedding coding vectors and the penalty risk point semantic embedding coding vectors corresponding to the cosine similarity greater than a preset threshold in the risk indicator-penalty risk point semantic association topological matrix to obtain the set of financial risk assessment point semantic embedding coding vectors.

[0030] In the specific implementation, this application uses cosine similarity as a metric to evaluate the semantic correlation between each risk indicator semantic embedding coding vector and each penalty risk point semantic embedding coding vector, thereby revealing the correlation between risk indicators and penalty risk points. Among them, cosine similarity is a method of measuring the cosine value of the angle between two vectors. The closer its value is to 1, the more similar the directions of the two vectors are, that is, the stronger the correlation between the risk indicator and the penalty risk point. Based on this, by pairing risk indicators with strong correlations with penalty risk points to construct a set of semantic embedding coding vectors of financial risk assessment points, it is possible to effectively identify and integrate the deep semantic connection between risk indicators and penalty risk points, providing more accurate and comprehensive data support for financial risk assessment.

[0031] In the above-mentioned financial risk warning method, the step S130 semantically encodes the financial business to obtain a financial business semantic coding vector. Considering that in financial risk warning, financial business is usually described in text form, such as loan applications, investment plans, transaction records, etc. Therefore, in order to enable the computer to understand and process this information, the present application further semantically encodes the financial business. In a specific example of the present application, a financial business semantic encoder based on the Bert model is used to semantically encode the financial business to obtain the financial business semantic coding vector. Those of ordinary skill in the art should know that the Bert model is a pre-trained language representation model, which has a strong transfer learning ability through pre-training on a large-scale general corpus and can be applied to a variety of natural language processing tasks. In addition, unlike the traditional unidirectional language model, the Bert model uses a bidirectional Transformer architecture for text semantic understanding, so that it can simultaneously consider the contextual information on the left and right sides of each word unit in the text, so as to better understand the meaning of the word in the sentence. Based on this, this application uses the Bert model to semantically encode financial business descriptions, which can fully capture the deep semantic information in the financial business text and better understand the complex financial terms and contextual relationships therein, thereby effectively reflecting the essential content of the financial business and providing a more accurate semantic basis for subsequent financial risk assessment.

[0032] In the above-mentioned financial risk warning method, the step S140 performs a quick matching-oriented semantic query encoding on the set of the financial business semantic coding vector and the financial risk assessment point semantic embedding coding vector to obtain a financial business-financial risk assessment point query response semantic coding vector. Among them, since the financial business semantic coding vector contains relevant information of the specific financial business, and the financial risk assessment point semantic embedding coding vector represents known risk indicators and penalty risk points. Therefore, in order to realize the intelligent identification and early warning of financial business risks, the present application further performs semantic query interaction processing on the set of the financial business semantic coding vector and the financial risk assessment point semantic embedding coding vector, so as to identify the risk assessment point associated with the financial business semantic coding vector from the set of the financial risk assessment point semantic embedding coding vector, thereby mining the potential risk information related to the financial business.

[0033] Figure 4 FIG. 1 is a flow chart of step S140 in the financial risk early warning method according to an embodiment of the present application. Figure 4As shown, the step S140 includes: S141, determining the positioning center financial risk assessment point semantic embedding coding vector of the set of financial risk assessment point semantic embedding coding vectors based on the semantic difference between the financial business semantic coding vector and each financial risk assessment point semantic embedding coding vector in the set of financial risk assessment point semantic embedding coding vectors; S142, determining the financial risk assessment point fine-grained matching search window based on the positioning center financial risk assessment point semantic embedding coding vector; S143, performing fine-grained semantic query response interactive coding on the financial business semantic coding vector and each financial risk assessment point semantic embedding coding vector in the financial risk assessment point fine-grained matching search window to obtain the financial business-financial risk assessment point query response semantic coding vector.

[0034] Specifically, the step S141 includes: calculating the cross entropy of the financial business semantic coding vector relative to each financial risk assessment point semantic embedding coding vector in the set of financial risk assessment point semantic embedding coding vectors to obtain a set of financial risk assessment point fast query positioning factors; using the financial risk assessment point semantic embedding coding vector corresponding to the minimum value in the set of financial risk assessment point fast query positioning factors as the positioning center financial risk assessment point semantic embedding coding vector, which is expressed by the formula: in, represents the set of semantic embedding encoding vectors of the financial risk assessment points, , , and They represent the first, second, and third in the set of semantic embedding coding vectors of the financial risk assessment point respectively. and The semantic embedding encoding vector of financial risk assessment points, is the number of feature vectors in the set of semantic embedding encoding vectors of the financial risk assessment point, Indicates the The semantic embedding encoding vector of the financial risk assessment point The eigenvalues ​​at the positions, represents the financial business semantic coding vector, Indicates the first The eigenvalues ​​at the positions, represents the logarithmic function with base 2, , , and Respectively indicate the Relative to the , , and stated The cross entropy of represents a set of quick query positioning factors for the financial risk assessment point, represents the semantic embedding coding vector of the financial risk assessment point corresponding to the minimum value in the set of the financial risk assessment point quick query positioning factors, Represents the semantic embedding encoding vector of the location center financial risk assessment point.

[0035] In order to quickly locate the risk assessment points related to financial business in the set of semantic embedding coding vectors of the financial risk assessment points, first, the present application calculates the cross entropy of the financial business semantic coding vector relative to the semantic embedding coding vectors of each financial risk assessment point to quantify the semantic distance and mismatch between the two, and generates a set of financial risk assessment point fast query positioning factors. Among them, cross entropy, as a key indicator to measure the difference between two probability distributions, can effectively guide the query process to move closer to the target area, providing an accurate basis for subsequent fast query positioning. Next, in order to quickly narrow the search scope of the semantic query matching to the most likely relevant area, the minimum value in the set of financial risk assessment point fast query positioning factors calculated above is further used to determine the best matching financial risk assessment point semantic embedding coding vector, so as to ensure that the initially screened financial risk assessment point semantic embedding coding vector has the highest semantic relevance to the financial business.

[0036] Specifically, in step S142, the vector of the center position of the financial risk assessment point fine-grained matching search window is the semantic embedding coding vector of the positioning center financial risk assessment point, and each financial risk assessment point semantic embedding coding vector in the financial risk assessment point fine-grained matching search window is defined as a fine-grained query financial risk assessment point semantic embedding coding vector. The determination process of the financial risk assessment point fine-grained matching search window can be expressed by the formula: in, represents the width of the fine-grained matching search window of the financial risk assessment point, represents the calculation of the norm of the vector, Indicates rounding down. Represents the semantic embedding encoding vector of the positioning center financial risk assessment point The position index in the set of semantic embedding encoding vectors of the financial risk assessment point, , , and querying the semantic embedding encoding vector of the financial risk assessment point for each fine-grained matching search window of the financial risk assessment point, A set of semantic embedding coding vectors of each fine-grained query financial risk assessment point in the financial risk assessment point fine-grained matching search window is obtained.

[0037] A fine-grained matching search window for financial risk assessment points is defined with the semantic embedding coding vector of the financial risk assessment point of the positioning center initially screened as the center. Among them, the semantic embedding coding vectors of each financial risk assessment point in the window are relatively close to the positioning center in the semantic feature space, and can be used to represent potential risk matching items of financial business. Based on this, the semantic embedding coding vectors of each financial risk assessment point in the window are further designated as fine-grained query financial risk assessment point semantic embedding coding vectors, as candidates for the next step of semantic interaction response coding, for performing more sophisticated local context perception and semantic query interaction operations. In this way, resources can be concentrated on a smaller and more targeted data subset for in-depth analysis, which improves the accuracy of semantic queries while ensuring computational efficiency.

[0038] Specifically, the step S143 includes: performing a linear transformation on the financial business semantic encoding vector to obtain a query vector and a value vector, and using each fine-grained query financial risk assessment point semantic embedding encoding vector in the financial risk assessment point fine-grained matching search window as a key vector, inputting the query vector, the value vector and the key vector into a fine-grained query encoding module based on a heterogeneous converter structure to obtain the financial business-financial risk assessment point query response semantic encoding vector, which is expressed as follows: in, and denote the query embedding matrix and the value embedding matrix respectively, and Represent different bias terms, and denote the query vector and value vector respectively, represents the first A fine-grained query financial risk assessment point semantic embedding encoding vector, represents matrix multiplication, represents the soft maximization function, represents the transpose of a vector, Indicates the A fine-grained query of the characteristic scale value of the semantic embedding encoding vector of the financial risk assessment point, Represents the financial business-financial risk assessment point query response semantic encoding vector.

[0039] By performing a linear transformation on the financial business semantic coding vector to convert it into a query vector and a value vector suitable for the attention mechanism, and combining the semantic embedding coding vectors of each fine-grained query financial risk assessment point in the search window as a key vector, a fine-grained query coding module based on a heterogeneous converter structure is used to simulate the attention interaction between queries, keys and values, and a multi-head self-attention layer is used to parallelly process the semantic associations between financial business and each financial risk assessment point to achieve deep semantic interaction fusion between financial business and each financial risk assessment point, and generate a financial business-financial risk assessment point query response semantic coding vector that integrates the semantic interaction information between financial business and each financial risk assessment point, thereby providing a more accurate semantic context representation for subsequent financial business risk assessment.

[0040] In the above-mentioned financial risk early warning method, the step S150 determines the financial risk analysis result based on the semantic coding vector of the query response of the financial business-financial risk assessment point. In a specific example of the present application, the semantic coding vector of the query response of the financial business-financial risk assessment point is input into a financial risk analysis module based on a classifier to obtain the financial risk analysis result, and the financial risk analysis result is used to indicate whether the financial business has financial risk. Specifically, the classifier learns the semantic interaction information between the financial business and each financial risk assessment point contained therein by performing multi-level perception and comprehensive analysis on the semantic coding vector of the query response of the financial business-financial risk assessment point, and makes a judgment on the financial risk based on this. Among them, the output result of the classifier is a binary classification, such as "there is financial risk" and "there is no financial risk". In this way, potential risk points in financial business can be effectively identified, and timely risk warnings and decision-making support can be provided to financial institutions, so that financial institutions can take corresponding risk control measures in a timely manner, such as adjusting loan conditions, increasing margin requirements or rejecting certain high-risk financial businesses, thereby effectively reducing potential financial losses and maintaining the stability of the financial market.

[0041] In a preferred example of the present application, considering that the set of the financial business semantic coding vector and the set of the financial risk assessment point semantic embedded coding vector respectively represent the text semantic coding features of the financial business and the text semantic coding features of the risk indicators / penalty risk points, when performing semantic query encoding, due to the differences in the rapid positioning mechanism caused by the differences in the distribution of the embedded coding features of the source data, the financial business-financial risk assessment point query response semantic coding vector will also have probabilistic convergence and divergence based on different dynamic semantic query properties, thereby affecting the accuracy of the financial risk analysis results obtained through the classifier-based financial risk analysis module.

[0042] Therefore, when the financial service-financial risk assessment point query response semantic coding vector is passed through a classifier-based financial risk analysis module, the financial service-financial risk assessment point query response semantic coding vector is first optimized, including the following steps: Determine the semantic encoding vector of the financial business-financial risk assessment point query response by using the semantic encoding distribution probability value of the financial business-financial risk assessment point query response of the financial risk analysis module based on the classifier , wherein the financial business-financial risk assessment point query response semantic encoding distribution probability value Indicates the probability that the financial business has financial risks; Each eigenvalue in the financial business-financial risk assessment point query response semantic coding vector is normalized to the maximum value to obtain the financial business-financial risk assessment point query response semantic coding probability eigenvalue ; The probability value of the semantic encoding distribution of the query response of the financial business-financial risk assessment point is subtracted from one As the index, calculate the semantic encoding probability feature value of each financial business-financial risk assessment point query response The power function of the first financial business-financial risk assessment point query response semantic encoding probability convergence eigenvalue ,in, It represents the maximum value normalization of the semantic encoding vector of the financial business-financial risk assessment point query response. The semantic encoding probability feature value of the query response of the financial business-financial risk assessment point at each location, Indicates the The corresponding first financial business-financial risk assessment point query response semantic encoding probability convergence eigenvalue; The query response semantic encoding distribution probability value of the financial business-financial risk assessment point Calculate a probability feature value for the semantic encoding of each query response of the financial business-financial risk assessment point for the index The power function of the difference is used to obtain the second financial business-financial risk assessment point query response semantic encoding probability convergence eigenvalue ,in, Indicates the The corresponding second financial business-financial risk assessment point query response semantic encoding probability convergence eigenvalue; Each of the financial business-financial risk assessment point query response semantic encoding probability feature value The probability value of the semantic encoding distribution of the query response of the financial business-financial risk assessment point is subtracted from one The difference is multiplied to obtain the first financial business-financial risk assessment point query response semantic encoding probability limit feature value ,in, Indicates the The corresponding first financial business-financial risk assessment point query response semantic encoding probability limit feature value; A semantically encoded probability feature value associated with each of the financial business-financial risk assessment point query responses The difference between the financial business-financial risk assessment point query response semantic encoding distribution probability value Multiply to obtain the second financial business-financial risk assessment point query response semantic encoding probability limit feature value ,in, Indicates the The corresponding second financial business-financial risk assessment point query response semantic encoding probability limit feature value; After multiplying the first financial business-financial risk assessment point query response semantic coding probability convergence eigenvalue and the second financial business-financial risk assessment point query response semantic coding probability convergence eigenvalue, further adding the first financial business-financial risk assessment point query response semantic coding probability limit eigenvalue and the second financial business-financial risk assessment point query response semantic coding probability limit eigenvalue to obtain the optimized eigenvalue of the financial business-financial risk assessment point query response semantic coding vector ,in, It indicates that the semantic encoding vector of the query response of the financial business-financial risk assessment point is optimized. The optimized eigenvalues ​​at each position.

[0043] By using the cross entropy form power series of the regression probability of the financial business-financial risk assessment point query response semantic coding vector as the probability distribution convergence limit, the regression probability convergence approximation is performed on the basis of the combination of the feature set distribution and the probability density distribution of the financial business-financial risk assessment point query response semantic coding vector, so as to avoid the conservative convergence control of the fatal potential divergence of the financial business-financial risk assessment point query response semantic coding vector, further use its probability distribution cross entropy as the objective function to guide the limited recovery strategy, so as to achieve the common agility of the convergence of the probability convergence and divergence of the feature set of the financial business-financial risk assessment point query response semantic coding vector to the probability density distribution space, and improve the accuracy of the financial risk analysis result obtained by the financial business-financial risk assessment point query response semantic coding vector through the classifier-based financial risk analysis module.

[0044] In summary, the financial risk warning method according to the embodiment of the present application is explained, which first obtains the latest financial regulatory documents and the latest financial penalty information, and extracts the risk indicator set and the penalty risk point set from them, and then uses the natural language processing technology based on deep learning to perform semantic embedding encoding on the risk indicator set and the penalty risk point set, and based on the correlation between the risk indicator and the penalty risk point, constructs a set of semantic embedding representations of financial risk assessment points, and then through rapid semantic query interaction between financial business and each financial risk assessment point, in order to dig out the potential risk information of financial business, so as to realize intelligent identification and warning of financial business risks. In this way, the intelligence level of financial risk warning can be effectively improved, the ability to identify financial business risks can be enhanced, and the compliance and security of financial business can be ensured.

[0045] Furthermore, the present application also provides a financial risk early warning system.

[0046] Figure 5 FIG. 1 is a block diagram of a financial risk early warning system according to an embodiment of the present application. Figure 5As shown, according to the financial risk early warning system 100 of the embodiment of the present application, it includes: a risk data semantic embedding coding module 110, which is used to perform semantic embedding coding on each risk indicator in the risk indicator set and each penalty risk point in the penalty risk point set to obtain a set of risk indicator semantic embedding coding vectors and a set of penalty risk point semantic embedding coding vectors; a risk data association merging module 120, which is used to perform set merging on the set of risk indicator semantic embedding coding vectors and the set of penalty risk point semantic embedding coding vectors to obtain a set of financial risk assessment point semantic embedding coding vectors; a financial business semantic coding module 130, which is used to perform semantic coding on financial business to obtain a financial business semantic coding vector; a fast matching semantic query coding module 140, which is used to perform fast matching-oriented semantic query coding on the set of financial business semantic coding vectors and the set of financial risk assessment point semantic embedding coding vectors to obtain a financial business-financial risk assessment point query response semantic coding vector; an analysis result generation module 150, which is used to determine the financial risk analysis result based on the financial business-financial risk assessment point query response semantic coding vector.

[0047] The specific operations of each module in the above financial risk early warning system have been referenced above. Figures 1 to 4 The financial risk early warning method has been introduced in detail in the description of the financial risk early warning method, and therefore, its repeated description will be omitted.

[0048] A storage medium stores a financial risk warning program, which implements a financial risk warning method when executed by a processor.

[0049] In addition to the above-mentioned methods and systems, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the financial risk warning method according to various embodiments of the present application described in the above-mentioned exemplary method section of this specification.

[0050] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0051] In addition, an embodiment of the present application may also be a computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor executes the steps of the financial risk warning method according to various embodiments of the present application described in the above exemplary method section of this specification.

[0052] The computer-readable storage medium may adopt any combination of one or more readable storage media. The readable storage medium may be a readable signal storage medium or a readable storage medium. The readable storage medium may, for example, include but is not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0053] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A financial risk early warning method, comprising: Obtain the latest financial regulatory documents and the latest financial penalty information; Parsing the latest financial regulatory documents to obtain a set of risk indicators; Parsing the latest financial penalty information to obtain a penalty risk point set; Based on the risk indicator set and the penalty risk point set, a risk warning is provided for the financial business, wherein based on the risk indicator set and the penalty risk point set, a risk warning is provided for the financial business, including: Semantically embedding and coding each risk indicator in the risk indicator set and each penalty risk point in the penalty risk point set to obtain a set of risk indicator semantic embedding coding vectors and a set of penalty risk point semantic embedding coding vectors; Merging the set of risk indicator semantic embedding coding vectors and the set of penalty risk point semantic embedding coding vectors to obtain a set of financial risk assessment point semantic embedding coding vectors; Performing semantic coding on the financial service to obtain a financial service semantic coding vector; Performing fast matching-oriented semantic query encoding on the set of the financial business semantic encoding vector and the financial risk assessment point semantic embedding encoding vector to obtain a financial business-financial risk assessment point query response semantic encoding vector; Based on the financial business-financial risk assessment point query response semantic coding vector, a financial risk analysis result is determined.

2. The financial risk early warning method according to claim 1, characterized in that: The set of risk indicator semantic embedding coding vectors and the set of penalty risk point semantic embedding coding vectors are combined to obtain a set of financial risk assessment point semantic embedding coding vectors, including: Calculating the cosine similarity between each risk indicator semantic embedding coding vector in the set of risk indicator semantic embedding coding vectors and each penalty risk point semantic embedding coding vector in the set of penalty risk point semantic embedding coding vectors to obtain a risk indicator-penalty risk point semantic association topological matrix; The risk indicator semantic embedding coding vectors and the penalty risk point semantic embedding coding vectors corresponding to the cosine similarity greater than a preset threshold in the risk indicator-penalty risk point semantic association topological matrix are cascaded to obtain a set of financial risk assessment point semantic embedding coding vectors.

3. The financial risk early warning method according to claim 2, characterized in that: Semantically encoding the financial service to obtain a financial service semantic encoding vector includes: The financial service is semantically encoded using a financial service semantic encoder based on the Bert model to obtain the financial service semantic encoding vector.

4. The financial risk early warning method according to claim 3, characterized in that: The method of performing fast matching-oriented semantic query encoding on the set of the financial business semantic encoding vector and the financial risk assessment point semantic embedding encoding vector to obtain a financial business-financial risk assessment point query response semantic encoding vector includes: Determine the positioning center financial risk assessment point semantic embedding coding vector of the set of financial risk assessment point semantic embedding coding vectors based on the semantic difference between the financial business semantic encoding vector and each financial risk assessment point semantic embedding coding vector in the set of financial risk assessment point semantic embedding coding vectors; Determining a fine-grained matching search window for financial risk assessment points based on the semantic embedding coding vector of the positioning center financial risk assessment point; Fine-grained semantic query response interactive encoding is performed on the financial business semantic encoding vector and each financial risk assessment point semantic embedding encoding vector in the financial risk assessment point fine-grained matching search window to obtain the financial business-financial risk assessment point query response semantic encoding vector.

5. The financial risk early warning method according to claim 4, characterized in that: Determining the positioning center financial risk assessment point semantic embedding coding vector of the set of financial risk assessment point semantic embedding coding vectors based on the semantic difference between the financial business semantic encoding vector and each financial risk assessment point semantic embedding coding vector in the set of financial risk assessment point semantic embedding coding vectors includes: Calculating the cross entropy of the financial business semantic encoding vector relative to each financial risk assessment point semantic embedding encoding vector in the set of financial risk assessment point semantic embedding encoding vectors to obtain a set of financial risk assessment point fast query positioning factors; The financial risk assessment point semantic embedding coding vector corresponding to the minimum value in the set of the financial risk assessment point quick query positioning factors is used as the positioning center financial risk assessment point semantic embedding coding vector.

6. The financial risk early warning method according to claim 5, characterized in that: The vector of the center position of the financial risk assessment point fine-grained matching search window is the semantic embedding coding vector of the positioned center financial risk assessment point, and the semantic embedding coding vectors of each financial risk assessment point in the financial risk assessment point fine-grained matching search window are defined as fine-grained query financial risk assessment point semantic embedding coding vectors.

7. The financial risk early warning method according to claim 6, characterized in that: The financial business semantic coding vector and each financial risk assessment point semantic embedding coding vector in the financial risk assessment point fine-grained matching search window are subjected to fine-grained semantic query response interactive coding to obtain the financial business-financial risk assessment point query response semantic coding vector, including: The financial business semantic coding vector is linearly transformed to obtain a query vector and a value vector, and each fine-grained query financial risk assessment point semantic embedding coding vector in the financial risk assessment point fine-grained matching search window is used as a key vector, and the query vector, the value vector and the key vector are input into a fine-grained query coding module based on a heterogeneous converter structure to obtain the financial business-financial risk assessment point query response semantic coding vector.

8. The financial risk early warning method according to claim 7, characterized in that: Determining a financial risk analysis result based on the financial business-financial risk assessment point query response semantic coding vector includes: The financial business-financial risk assessment point query response semantic encoding vector is input into a classifier-based financial risk analysis module to obtain the financial risk analysis result, and the financial risk analysis result is used to indicate whether there is a financial risk in the financial business.

9. A financial risk early warning system, characterized in that: include: A risk data semantic embedding coding module is used to perform semantic embedding coding on each risk indicator in the risk indicator set and each penalty risk point in the penalty risk point set to obtain a set of risk indicator semantic embedding coding vectors and a set of penalty risk point semantic embedding coding vectors; A risk data association merging module, used for merging the set of risk indicator semantic embedding coding vectors and the set of penalty risk point semantic embedding coding vectors to obtain a set of financial risk assessment point semantic embedding coding vectors; A financial service semantic coding module, used for semantically coding financial services to obtain financial service semantic coding vectors; A fast matching semantic query encoding module, used for performing fast matching-oriented semantic query encoding on the set of the financial business semantic encoding vector and the financial risk assessment point semantic embedding encoding vector to obtain a financial business-financial risk assessment point query response semantic encoding vector; The analysis result generating module is used to determine the financial risk analysis result based on the financial business-financial risk assessment point query response semantic coding vector.

10. A storage medium, characterized in that: The storage medium stores a financial risk warning program, and when the financial risk warning program is executed by the processor, the financial risk warning method according to any one of claims 1 to 8 is implemented.

Images