Early warning analysis method, system and readable storage medium for financial data

Abstract

The present disclosure relates to a warning analysis method, system and readable storage medium for financial data. It includes: semantically encoding the personal information, credit record and financial status in the acquired financial related data respectively, and then performing semantic co-space mapping to obtain the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector; then, extracting the characteristics of the association pattern between financial data from the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector, and inputting them into the category-related attention allocation network based on the gating mechanism to obtain the semantic association enhancement feature map between the financial data of the borrower and the borrower, and then determining whether the credit risk of the borrower exceeds the predetermined risk level. In this way, the repayment ability and default risk of the borrower can be predicted, so as to evaluate and warn the credit risk of the borrower, and provide more accurate credit decision support for financial institutions.

Description

Technical Field

[0001] The present disclosure relates to the technical field of early warning analysis, and in particular, to an early warning analysis method, system and readable storage medium for financial data. Background Art

[0002] With the rapid development of financial markets and the increasing diversification of financial products, financial institutions are facing more complex credit risk management challenges. Financial institutions need to assess the credit risk of borrowers when lending to ensure loan safety and minimize the risk of default. Traditionally, financial institutions usually rely on manual review and statistical model-based methods to assess the credit risk of borrowers. Although these methods are effective to a certain extent, they have some defects when faced with large-scale data and complex risk situations. For example, manual review is easily affected by human subjective factors, has low accuracy and efficiency, and is difficult to cope with large-scale financial data. In addition, traditional methods are usually based on static statistical models and cannot fully adapt to the dynamic changes in financial markets and borrowers' conditions, resulting in challenges to the timeliness and accuracy of the model.

[0003] Therefore, an optimized early warning analysis solution for financial data is desired. Summary of the invention

[0004] This summary is provided to introduce concepts in a brief form that will be described in detail in the detailed description below. This summary is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0005] In a first aspect, the present disclosure provides an early warning analysis method for financial data, the method comprising: obtaining financial-related data of a borrower, wherein the financial-related data comprises personal information, credit record and financial status; semantically encoding the personal information, credit record and financial status in the financial-related data respectively to obtain a personal information semantic encoding vector, a credit record semantic encoding vector and a financial status semantic encoding vector; semantically co-space mapping the personal information semantic encoding vector, the credit record semantic encoding vector and the financial status semantic encoding vector to obtain a personal information mapping semantic vector, a credit record mapping semantic encoding vector and a financial status mapping semantic encoding vector; extracting the features of the association pattern between financial data from the personal information mapping semantic vector, the credit record mapping semantic encoding vector and the financial status mapping semantic encoding vector to obtain a semantic association feature map between the financial data of the borrower and the lender; inputting the semantic association feature map between the financial data of the borrower and the lender into a category-related attention allocation network based on a gating mechanism to obtain a semantic association enhanced feature map between the financial data of the borrower and the lender; and determining whether the credit risk of the borrower and the lender exceeds a predetermined risk level based on the semantic association enhanced feature map between the financial data of the borrower and the lender.

[0006] Optionally, the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector are semantically co-space mapped to obtain the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector, including: passing the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector respectively through a semantic co-space mapper based on a fully connected layer to obtain the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector.

[0007] Optionally, the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector are subjected to financial data association pattern feature extraction to obtain a semantic association feature graph between the borrower's financial data, including: arranging the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector into a semantic association matrix and then passing them through a financial data association pattern feature extractor based on a void convolutional neural network model to obtain the semantic association feature graph between the borrower's financial data.

[0008] Optionally, the semantic association feature map between the financial data of the borrowers and lenders is input into a category-dependent attention allocation network based on a gating mechanism to obtain an enhanced semantic association feature map between the financial data of the borrowers and lenders, including: performing a maximum-based global pooling process, a random-value-based global pooling process, and an average-based global pooling process on the semantic association feature map between the financial data of the borrowers and lenders to obtain a global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and lenders, a global mean pooling feature vector of the semantic association features between the financial data of the borrowers and lenders, and a global random value pooling feature vector of the semantic association features between the financial data of the borrowers and lenders; calculating the position-based sum of the global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and lenders, the global mean pooling feature vector of the semantic association features between the financial data of the borrowers and lenders, and the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and lenders; The weights are summed to obtain a multi-scale pooled representation vector of semantic association features between financial data of lenders and borrowers; the multi-scale pooled representation vector of semantic association features between financial data of lenders and borrowers is subjected to channel-by-channel semantic feature interaction and feature activation based on a fully connected layer to obtain a channel semantic association feature vector between financial data of lenders and borrowers; the channel semantic association feature vector between financial data of lenders and borrowers is input into a gating unit to perform masking processing on the feature values ​​of each position in the feature vector of channel semantic association between financial data of lenders and borrowers and then perform a normalization operation to obtain a normalized semantic weight vector between financial data of lenders and borrowers; the feature values ​​of each position in the normalized semantic weight vector between financial data of lenders and borrowers are used as weighting coefficients to perform weighted optimization on each feature matrix along the channel dimension in the semantic association enhanced feature map between financial data of lenders and borrowers to obtain the semantic association enhanced feature map between financial data of lenders and borrowers.

[0009] Optionally, calculating the position-weighted sum of the global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and lenders, the global mean pooling feature vector of the semantic association features between the financial data of the borrowers and lenders, and the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and lenders to obtain a multi-scale pooling representation vector of the semantic association features between the financial data of the borrowers and lenders, including: calculating the position-weighted sum of the global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and lenders and a first preset trainable weight value to obtain a weight-modulated global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and lenders; calculating the position-weighted sum of the global mean pooling feature vector of the semantic association features between the financial data of the borrowers and lenders and a second preset trainable weight value to obtain a weight-modulated global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and lenders; The method comprises the following steps: performing positional multiplication of the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders to obtain a global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation; calculating the positional multiplication of the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders and the lenders with a third preset trainable weight value to obtain a global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation; adding the global maximum value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation, the global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation, and the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation to obtain a multi-scale pooling representation vector of the semantic association features between the financial data of the borrowers and the lenders.

[0010] Optionally, the multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders is subjected to channel-by-channel semantic feature interaction and feature activation based on a fully connected layer to obtain the channel semantic association feature vector between the financial data of the borrowers and the lenders, including: calculating the matrix multiplication of the multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders and then performing vector addition with the bias vector to obtain the bias-adjusted multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders; and the bias-adjusted multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders is obtained by Activation processing is performed in the activation function to obtain the channel semantic association feature vector between the borrower's financial data.

[0011] Optionally, the channel semantic association feature vector between the financial data of the lenders and the borrowers is input into a gating unit to perform masking processing on each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers, and then perform normalization operation to obtain a normalized semantic weight vector between the financial data of the lenders and the borrowers, including: based on a gating mechanism, masking processing is performed on each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers to obtain a masked channel semantic association feature vector between the financial data of the lenders and the borrowers, wherein, in response to each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers being greater than or equal to a predetermined threshold, each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers takes an original value, otherwise it is set to zero; each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers is normalized to obtain a normalized semantic weight vector between the financial data of the lenders and the borrowers.

[0012] Optionally, based on the semantic association enhanced feature graph between the borrower's financial data, determining whether the borrower's credit risk exceeds a predetermined risk level includes: passing the semantic association enhanced feature graph between the borrower's financial data through a classifier-based early warning analyzer to obtain an early warning analysis result, wherein the early warning analysis result is used to indicate whether the borrower's credit risk exceeds a predetermined risk level.

[0013] In a second aspect, the present disclosure provides an early warning analysis system for financial data, the system comprising: a financial-related data acquisition module, used to acquire financial-related data of a borrower, wherein the financial-related data includes personal information, credit records and financial status; a semantic encoding module, used to semantically encode the personal information, credit records and financial status in the financial-related data to obtain a personal information semantic encoding vector, a credit record semantic encoding vector and a financial status semantic encoding vector; a semantic co-space mapping module, used to semantically co-space map the personal information semantic encoding vector, the credit record semantic encoding vector and the financial status semantic encoding vector to obtain a personal information mapping semantic vector, a credit record mapping semantic vector and a financial status mapping semantic vector. A semantic coding vector and a financial status mapping semantic coding vector; a financial data association pattern feature extraction module, used to extract the financial data association pattern features from the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector to obtain a semantic association feature map between the borrower's financial data; a category-related attention allocation module, used to input the semantic association feature map between the borrower's financial data into a category-related attention allocation network based on a gating mechanism to obtain a semantic association enhanced feature map between the borrower's financial data; a credit risk risk level determination module, used to determine whether the borrower's credit risk exceeds a predetermined risk level based on the semantic association enhanced feature map between the borrower's financial data.

[0014] In a third aspect, the present disclosure provides a readable storage medium having a computer program stored thereon, which, when executed by a processing device, implements the steps of the early warning analysis method for financial data.

[0015] The above technical solution is adopted to semantically encode the personal information, credit record and financial status in the acquired financial related data respectively and then perform semantic co-space mapping to obtain the personal information mapping semantic vector, credit record mapping semantic coding vector and financial status mapping semantic coding vector; then, the personal information mapping semantic vector, credit record mapping semantic coding vector and financial status mapping semantic coding vector are used to extract the characteristics of the association pattern between financial data, and input into the category-related attention allocation network based on the gating mechanism to obtain the semantic association enhanced feature map between the financial data of the borrower and the borrower, thereby determining whether the credit risk of the borrower exceeds the predetermined risk level. In this way, the repayment ability and default risk of the borrower can be predicted, so as to evaluate and warn the credit risk of the borrower, and provide more accurate credit decision support for financial institutions.

[0016] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale.

[0018] In the attached picture: Figure 1 The present invention is a flowchart of a method for early warning analysis of financial data according to an exemplary embodiment.

[0019] Figure 2 The present invention is a block diagram of a system for early warning analysis of financial data according to an exemplary embodiment.

[0020] Figure 3 It is a block diagram of an electronic device according to an exemplary embodiment.

[0021] Figure 4 It is an application scenario diagram of an early warning analysis method for financial data according to an exemplary embodiment. DETAILED DESCRIPTION

[0022] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0023] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0024] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0025] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0026] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0027] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0028] In order to solve the above problems, the present disclosure provides a warning analysis method, system and readable storage medium for financial data, which respectively performs semantic coding on the personal information, credit record and financial status in the acquired financial related data and then performs semantic co-space mapping to obtain the personal information mapping semantic vector, credit record mapping semantic coding vector and financial status mapping semantic coding vector; then, the personal information mapping semantic vector, credit record mapping semantic coding vector and financial status mapping semantic coding vector are used to extract the characteristics of the association pattern between financial data, and input into the category-related attention allocation network based on the gating mechanism to obtain the semantic association enhancement feature map between the financial data of the borrower and the borrower, thereby determining whether the credit risk of the borrower exceeds the predetermined risk level. In this way, the repayment ability and default risk of the borrower can be predicted, thereby evaluating and warning the credit risk of the borrower, and providing more accurate credit decision support for financial institutions.

[0029] The specific embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0030] Figure 1 is a flow chart of a method for early warning analysis of financial data according to an exemplary embodiment. Figure 1 As shown, the method includes: step 101, obtaining the borrower's financial related data, wherein the financial related data includes personal information, credit record and financial status.

[0031] Step 102: semantically encode the personal information, credit record and financial status in the financial-related data respectively to obtain a personal information semantic encoding vector, a credit record semantic encoding vector and a financial status semantic encoding vector.

[0032] Step 103: semantically co-space map the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector to obtain a personal information mapping semantic coding vector, a credit record mapping semantic coding vector and a financial status mapping semantic coding vector.

[0033] Step 104: extract the features of the association pattern between financial data from the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector to obtain a semantic association feature graph between the borrower's financial data.

[0034] Step 105: Input the semantic association feature map between the borrower's financial data into a category-dependent attention allocation network based on a gating mechanism to obtain a semantic association enhanced feature map between the borrower's financial data.

[0035] Step 106: Determine whether the credit risk of the borrower exceeds a predetermined risk level based on the semantic association enhanced feature graph between the borrower's financial data.

[0036] In response to the above technical problems, the technical concept of the present invention is to analyze the financial-related data of the borrower through financial data processing technology, which not only includes the borrower's basic information, such as age, gender, occupation, etc., but also includes more complex data such as credit records and financial status, so as to predict the borrower's repayment ability and default risk, thereby evaluating and warning the borrower's credit risk, and providing financial institutions with more accurate credit decision support.

[0037] Specifically, in the technical solution of the present disclosure, first, the financial data of the borrower is obtained, wherein the financial data includes personal information, credit record and financial status. It should be understood that personal information includes basic information of the borrower, such as age, gender, occupation, etc., which can provide basic semantic information about the identity and background of the borrower, and help to understand the personal characteristics and background of the borrower. Credit record is an important data reflecting the credit history and credit behavior of the borrower, which can reflect more complex credit information such as the borrower's past credit performance, loan record, overdue situation, etc., so as to help assess the credit risk of the borrower. The financial status involves the financial health of the borrower, including information on income, expenditure, assets, liabilities, etc., which can help understand the financial status of the borrower and assess its repayment ability and debt repayment ability.

[0038] Based on this, in the technical solution of the present disclosure, in order to be able to perform semantic analysis and encoding on the financial-related data of these borrowers to capture the semantics of the financial data related to the borrowers, so as to facilitate the subsequent assessment of the credit risk of the borrowers, in the technical solution of the present disclosure, it is necessary to perform semantic encoding on the personal information, credit record and financial status in the financial-related data to obtain the personal information semantic encoding vector, credit record semantic encoding vector and financial status semantic encoding vector. Through semantic encoding, the semantic features of the borrower's personal information, credit record and financial status can be captured respectively. These semantic features contain basic semantic information about the borrower's identity and background, semantic information about the borrower's credit history and credit behavior, and information about the borrower's financial health status, which helps to more fully and accurately understand the semantics of the borrower's financial data and provide a basis for the subsequent credit risk assessment of the borrower.

[0039] Then, considering that although semantic coding of the borrower's financial-related data can extract the borrower's personal information semantic features, credit record semantic features and financial status semantic features, the source data of these data may be different data sources and different storage formats, resulting in the semantic features of these semantically coded financial-related data not being in the same semantic space. Based on this, in the technical solution disclosed in the present invention, the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector are further passed through a semantic co-space mapper based on a fully connected layer to obtain a personal information mapping semantic vector, a credit record mapping semantic coding vector and a financial status mapping semantic coding vector. Through the processing of the semantic co-space mapper based on a fully connected layer, the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector can be mapped to the same semantic space, which is conducive to the subsequent realization of information fusion and interaction between semantic coding features of different types of financial-related data, helping the model to better understand the correlation between different types of financial-related data and improve the comprehensive grasp of the overall situation of the borrower. It is worth mentioning that in another specific example of the present disclosure, the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector can also be mapped into a hash space, and the comparison is not limited to the present disclosure.

[0040] In one embodiment of the present disclosure, the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector are semantically co-space mapped to obtain the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector, including: passing the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector respectively through a semantic co-space mapper based on a fully connected layer to obtain the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector.

[0041] Next, considering that there is a correlation relationship between the semantic vector of the personal information mapping, the semantic coding vector of the credit record mapping and the semantic coding vector of the financial status mapping after the semantic common space mapping, this correlation relationship is of great significance for the credit risk assessment of the borrower. Based on this, in order to more deeply explore the semantic association and interaction information between different types of financial-related data of the borrower, thereby improving the comprehensive assessment ability of the borrower's credit risk, in the technical solution of the present disclosure, the semantic vector of the personal information mapping, the semantic coding vector of the credit record mapping and the semantic coding vector of the financial status mapping are further arranged into a semantic association matrix and then passed through a financial data association pattern feature extractor based on a hollow convolutional neural network model to obtain a semantic association feature map between the borrower's financial data. Specifically, since personal information, credit records and financial status are data from different fields, there may be complex associations and interactions between them. Arranging them into a semantic association matrix and processing them through a convolutional neural network model helps to integrate the information of different types of data and explore the potential associations between various types of financial data. In particular, the atrous convolutional neural network model can capture the semantic association features between each position in the semantic association matrix. The obtained semantic association feature graph between the borrower's financial data can intuitively display the degree of association between the semantics of different financial data, providing a more intuitive reference for subsequent risk assessment and decision-making.

[0042] In one embodiment of the present disclosure, the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector are subjected to financial data association pattern feature extraction to obtain a semantic association feature graph between the borrower's financial data, including: arranging the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector into a semantic association matrix and then passing through a financial data association pattern feature extractor based on a void convolutional neural network model to obtain the semantic association feature graph between the borrower's financial data.

[0043] Furthermore, since the semantic association feature graph between the borrower's financial data contains the semantic features of association between the borrower's personal information, credit record and financial status, but this semantic feature is of importance, some of which have little correlation with the judgment of credit risk. Therefore, in order to screen and enhance the important semantic features of financial data related to the borrower's credit risk, thereby improving the accuracy and pertinence of risk assessment, in the technical solution disclosed in the present invention, the semantic association feature graph between the borrower's financial data is further input into the category-related attention allocation network based on the gating mechanism to obtain the semantic association enhancement feature graph between the borrower's financial data. It should be understood that through the processing of the category-related attention allocation network based on the gating mechanism, the semantic association feature graph between financial data can be screened to extract important features with high correlation with the borrower's credit risk. This helps to reduce the amount of data information and focus on the most relevant and informative features for the credit risk assessment task. By screening and enhancing the important semantic features of financial data related to the borrower's credit risk, the model can more accurately identify potential risk factors and improve the accuracy and pertinence of risk assessment. This helps financial institutions to more accurately assess the credit risk of borrowers and reduce the risk of default.

[0044] In one embodiment of the present disclosure, the semantic association feature map between the financial data of the borrowers and the lenders is input into a category-dependent attention allocation network based on a gating mechanism to obtain a semantic association enhanced feature map between the financial data of the borrowers and the lenders, including: performing a maximum-based global pooling process, a random-value-based global pooling process, and an average-based global pooling process on the semantic association feature map between the financial data of the borrowers and the lenders to obtain a global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, a global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, and a global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders; calculating the relationship between the global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, the global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, and the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders. The multi-scale pooling representation vector of the semantic association features between the financial data of the borrowers and the lenders is obtained by weighting the sum by position; the multi-scale pooling representation vector of the semantic association features between the financial data of the borrowers and the lenders is subjected to channel-by-channel semantic feature interaction and feature activation based on a fully connected layer to obtain a channel semantic association feature vector between the financial data of the borrowers and the lenders; the channel semantic association feature vector between the financial data of the borrowers and the lenders is input into a gating unit to mask the feature values ​​of each position in the channel semantic association feature vector between the financial data of the borrowers and the lenders and then normalize the feature values ​​to obtain a normalized semantic weight vector between the financial data of the borrowers and the lenders; the feature values ​​of each position in the normalized semantic weight vector between the financial data of the borrowers and the lenders are used as weighting coefficients to perform weighted optimization on each feature matrix along the channel dimension in the semantic association enhanced feature map between the financial data of the borrowers and the lenders to obtain the semantic association enhanced feature map between the financial data of the borrowers and the lenders.

[0045] Further, in one embodiment of the present disclosure, the position-weighted sum of the global maximum pooling feature vector of the semantic association feature between the financial data of the borrowers and the lenders, the global mean pooling feature vector of the semantic association feature between the financial data of the borrowers and the lenders, and the global random value pooling feature vector of the semantic association feature between the financial data of the borrowers and the lenders is calculated to obtain a multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders, including: calculating the position-weighted sum of the global maximum pooling feature vector of the semantic association feature between the financial data of the borrowers and the lenders and the first preset trainable weight value to obtain the weight-modulated global maximum pooling feature vector of the semantic association feature between the financial data of the borrowers and the lenders; calculating the position-weighted sum of the global mean pooling feature vector of the semantic association feature between the financial data of the borrowers and the second ... lenders The weight values ​​are multiplied by the position to obtain a global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation; the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders is multiplied by the position to obtain a global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation; the global maximum value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation, the global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation and the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders are multiplied by the position to obtain a multi-scale pooling representation vector of the semantic association features between the financial data of the borrowers and the lenders.

[0046] Furthermore, in one embodiment of the present disclosure, the multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders is subjected to channel-by-channel semantic feature interaction and feature activation based on a fully connected layer to obtain the channel semantic association feature vector between the financial data of the borrowers and the lenders, including: calculating the matrix multiplication of the multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders and then adding the vector to the bias vector to obtain the bias-adjusted multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders; and applying the bias-adjusted multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders through Activation processing is performed in the activation function to obtain the channel semantic association feature vector between the borrower's financial data.

[0047] Furthermore, in one embodiment of the present disclosure, the channel semantic association feature vector between the financial data of the lenders and the borrowers is input into a gating unit to perform masking processing on each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers, and then perform normalization operation to obtain a normalized semantic weight vector between the financial data of the lenders and the borrowers, including: based on a gating mechanism, masking processing is performed on each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers to obtain a masked channel semantic association feature vector between the financial data of the lenders and the borrowers, wherein, in response to each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers being greater than or equal to a predetermined threshold, each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers takes an original value, otherwise it is set to zero; each position feature value in the channel semantic association feature vector between the financial data of the lenders and the borrowers is normalized to obtain a normalized semantic weight vector between the financial data of the lenders and the borrowers.

[0048] Specifically, the semantic association feature map between the financial data of the borrowers and lenders is input into the category-related attention allocation network based on the gating mechanism and processed by the following category-related attention enhancement formula to obtain the semantic association enhancement feature map between the financial data of the borrowers and lenders; wherein the category-related attention enhancement formula is: ;in, is a semantic association feature graph between the borrower's financial data, , and They are global mean pooling, maximum value pooling and random value pooling. , and are preset trainable weight values, is the multi-scale pooling representation vector of semantic association features between borrowers’ financial data, and are the weight matrix and bias vector of the convolutional layer, respectively. for Activation function, is the hidden semantic association feature vector between the borrower’s financial data, is the hidden semantic association feature vector between the masked borrower and lender financial data, is the hidden semantic association feature vector between the masked borrower financial data. The eigenvalues ​​at the positions, is the semantic weight vector between normalized borrower and lender financial data, It means point multiplication by position. It is a semantic association enhanced feature graph between the borrower's financial data.

[0049] Then, the semantic association enhanced feature graph between the borrower's financial data is passed through a classifier-based early warning analyzer to obtain an early warning analysis result, which is used to indicate whether the borrower's credit risk exceeds a predetermined risk level. In other words, the semantic association features between the borrower's financial data after the significant features are enhanced are used for classification processing to predict the borrower's repayment ability and default risk, thereby evaluating and warning the borrower's credit risk, and providing more accurate credit decision support for financial institutions.

[0050] In one embodiment of the present disclosure, based on the semantic association enhanced feature graph between the borrower's financial data, determining whether the borrower's credit risk exceeds a predetermined risk level includes: passing the semantic association enhanced feature graph between the borrower's financial data through a classifier-based early warning analyzer to obtain an early warning analysis result, wherein the early warning analysis result is used to indicate whether the borrower's credit risk exceeds a predetermined risk level.

[0051] In the technical solution of the present application, when the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector are respectively passed through a semantic co-space mapper based on a fully connected layer, the semantic co-space mapper based on a fully connected layer uses the weight matrix of the fully connected layer to perform feature linear modulation on the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector so that the feature distributions of the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector have structural consistency, that is, the feature dimensions and structural correspondence of the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector are relatively good. In this way, after the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector are arranged into a semantic association matrix and passed through the financial data association pattern feature extractor based on the void convolutional neural network model, the financial data association pattern feature extractor based on the void convolutional neural network model can be used to capture the local semantic association features of personal information, credit records and financial status in the semantic common space, and considering that the local semantic association features of personal information, credit records and financial status have different contributions to the final classification judgment, therefore, in the technical solution of the present application, a category-related attention allocation network based on a gating mechanism is further used to enhance the local feature significance of the semantic association feature map between the borrower's financial data along the channel dimension to improve the distinguishability of each local semantic association feature of the semantic association feature map between the borrower's financial data along the channel dimension. However, when the category-dependent attention allocation network based on the gating mechanism is used to enhance the discrimination of local semantic association features, the overall semantic association enhancement feature map between the financial data of borrowers and lenders will also have local overflow features (for example, some local semantic features are over-enhanced and some local semantic features are incorrectly enhanced), so that the semantic association enhancement feature map between the financial data of borrowers and lenders has a multimodal diversified set representation distribution along the channel dimension. Therefore, it is expected to improve the balance between the accuracy and completeness of the class regression mapping of the semantic association enhancement feature map between the financial data of borrowers and lenders when performing classification regression for a classifier-based warning analyzer, thereby improving the accuracy of the obtained warning analysis results.

[0052] Based on this, in a preferred example, inputting the semantic association enhancement feature map between the financial data of lenders and borrowers into a warning analyzer based on a classifier to obtain a warning analysis result specifically includes: determining the maximum eigenvalue of the semantic association enhancement between the financial data of lenders and borrowers and the minimum eigenvalue of the semantic association enhancement between the financial data of lenders and borrowers of the semantic association enhancement feature map, and calculating the mean value of the semantic association enhancement between the financial data of lenders and borrowers and the standard deviation of the semantic association enhancement between the financial data of lenders and borrowers of the feature set of the semantic association enhancement feature map; calculating the quotient of the mean value of the semantic association enhancement between the financial data of lenders and borrowers and the standard deviation of the semantic association enhancement between the financial data of lenders and borrowers to obtain a statistical normalized value of the semantic association enhancement between the financial data of lenders and borrowers; calculating the inverse of each eigenvalue of the semantic association enhancement feature map between the financial data of lenders and borrowers, and comparing it with the semantic association enhancement between the financial data of lenders and borrowers. The method comprises the following steps: multiplying the difference between the maximum eigenvalue of the semantic association enhancement and the minimum eigenvalue of the semantic association enhancement between the financial data of the borrowers and the lenders, and then subtracting the result from the statistical normalization value of the semantic association enhancement between the financial data of the borrowers and the lenders to obtain a distribution approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders; calculating an exponential function with a natural constant as the base, with each eigenvalue of the distribution approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders as the exponent, and adding the result to the statistical normalization value of the semantic association enhancement between the financial data of the borrowers and the lenders to obtain a probability approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders; calculating the logarithm of the absolute value of each eigenvalue of the probability approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders to obtain an optimized feature map of the semantic association enhancement between the financial data of the borrowers and the lenders; and inputting the optimized feature map of the semantic association enhancement between the financial data of the borrowers and the lenders into a warning analyzer based on a classifier to obtain a warning analysis result.

[0053] Among them, the semantic correlation enhanced feature graph between the borrower's financial data is denoted as The optimization expression is: ;in, represents the semantic association enhanced feature graph between the borrower's financial data, , , and They are the feature sets of the semantic association enhanced feature graphs between the borrower's financial data The mean, standard deviation, minimum and maximum values ​​of It means point multiplication by position. It means adding by position. It means subtracting by position. Indicates the calculation of the inverse of each eigenvalue of the feature map, Represents the optimized semantic correlation enhanced feature graph between borrowers’ financial data.

[0054] That is, by performing random statistical normalization on the diversified feature set of the semantic association enhanced feature map between the borrower's financial data, the confidence space constructed based on the overall feature value of the semantic association enhanced feature map between the borrower's financial data is approximated by the standardized continuous probability density distribution relative to the response hypothesis test of each feature value of the semantic association enhanced feature map between the borrower's financial data, thereby establishing the target reachability of the discretized feature distribution of the semantic association enhanced feature map between the borrower's financial data to the unified regression target, so as to achieve the balanced executableness between the mapping accuracy and the mapping completeness in the classification regression process based on the discretized feature distribution of the semantic association enhanced feature map between the borrower's financial data, and improve the accuracy of the early warning analysis results obtained by the early warning analyzer based on the input of the semantic association enhanced feature map between the borrower's financial data. In this way, the repayment ability and default risk of the borrower can be predicted more accurately, so as to evaluate and warn the credit risk of the borrower, and provide more accurate credit decision support for financial institutions.

[0055] To sum up, the above scheme is adopted to analyze the financial-related data of the borrower through financial data processing technology, which not only includes the basic information of the borrower, such as age, gender, occupation, etc., but also includes more complex data such as credit records and financial status, so as to predict the borrower's repayment ability and default risk, thereby evaluating and warning the borrower's credit risk, and providing financial institutions with more accurate credit decision support.

[0056] Figure 2 FIG. 1 is a block diagram of a system for early warning analysis of financial data according to an exemplary embodiment. Figure 2As shown, the system 200 includes: a financial-related data acquisition module 201, which is used to acquire the financial-related data of the borrower, wherein the financial-related data includes personal information, credit records and financial status; a semantic encoding module 202, which is used to semantically encode the personal information, credit records and financial status in the financial-related data to obtain a personal information semantic encoding vector, a credit record semantic encoding vector and a financial status semantic encoding vector; a semantic co-space mapping module 203, which is used to semantically co-space map the personal information semantic encoding vector, the credit record semantic encoding vector and the financial status semantic encoding vector to obtain a personal information mapping semantic vector, a credit record mapping semantic encoding vector and a financial status semantic encoding vector. mapping semantic coding vector; a financial data association pattern feature extraction module 204, used to extract the financial data association pattern features from the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector to obtain a semantic association feature map between the borrower's financial data; a category-related attention allocation module 205, used to input the semantic association feature map between the borrower's financial data into a category-related attention allocation network based on a gating mechanism to obtain a semantic association enhanced feature map between the borrower's financial data; a credit risk risk level determination module 206, used to determine whether the borrower's credit risk exceeds a predetermined risk level based on the semantic association enhanced feature map between the borrower's financial data.

[0057] In one embodiment of the present disclosure, a readable storage medium is provided, on which a computer program is stored, characterized in that when the program is executed by a processing device, the steps of the early warning analysis method for financial data are implemented.

[0058] Reference below Figure 3 , which shows a schematic diagram of the structure of an electronic device 600 suitable for implementing the embodiment of the present disclosure. The terminal device in the embodiment of the present disclosure may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0059] like Figure 3As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0060] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0061] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0062] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer 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. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0063] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0064] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0065] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0066] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0067] The modules involved in the embodiments described in the present disclosure may be implemented by software or hardware. The name of a module does not limit the module itself in some cases. For example, a test parameter acquisition module may also be described as a "module for acquiring device test parameters corresponding to a target device".

[0068] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0069] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer 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 foregoing.

[0070] Figure 4 FIG. 1 is an application scenario diagram of an early warning analysis method for financial data according to an exemplary embodiment. Figure 4 As shown, in this application scenario, first, the borrower's financial data is obtained, wherein the financial data includes personal information (for example, Figure 4 C1 as indicated in the above), credit records (e.g. Figure 4 C2) and financial status (e.g. Figure 4 Then, the acquired personal information, credit record and financial status are input into a server that has deployed an early warning analysis algorithm for financial data (for example, Figure 4 In S) shown in , the server is capable of processing the personal information, the credit record and the financial status based on an early warning analysis algorithm for financial data to determine whether the borrower's credit risk exceeds a predetermined risk level.

[0071] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

[0072] In addition, although each operation is described in a specific order, this should not be understood as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0073] Although the subject matter has been described in language specific to structural features and / or method logic actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims. Regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.

Claims

1. A method for early warning analysis of financial data, characterized in that: include: Obtain financial data related to the borrower, wherein the financial data includes personal information, credit record and financial status; semantically encode the personal information, credit record and financial status in the financial data respectively to obtain a personal information semantic encoding vector, a credit record semantic encoding vector and a financial status semantic encoding vector; perform semantic co-space mapping on the personal information semantic encoding vector, the credit record semantic encoding vector and the financial status semantic encoding vector to obtain a personal information mapping semantic vector, a credit record mapping semantic encoding vector and a financial status mapping semantic encoding vector; extract the characteristics of the association pattern between financial data from the personal information mapping semantic vector, the credit record mapping semantic encoding vector and the financial status mapping semantic encoding vector to obtain a semantic association feature map between the financial data of the borrower; input the semantic association feature map between the financial data of the borrower into a category-dependent attention allocation network based on a gating mechanism to obtain a semantic association enhanced feature map between the financial data of the borrower; determine whether the credit risk of the borrower exceeds a predetermined risk level based on the semantic association enhanced feature map between the financial data of the borrower; The method comprises inputting the semantic association feature map between the financial data of the borrowers and the lenders into a category-dependent attention allocation network based on a gating mechanism to obtain an enhanced semantic association feature map between the financial data of the borrowers and the lenders, including: performing a maximum-based global pooling process, a random-value-based global pooling process and an average-based global pooling process on the semantic association feature map between the financial data of the borrowers and the lenders to obtain a global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, a global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, and a global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders; and calculating the position-based sum of the global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, the global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, and the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders. The weights are summed to obtain a multi-scale pooling representation vector of semantic association features between financial data of borrowers and lenders; the multi-scale pooling representation vector of semantic association features between financial data of borrowers and lenders is subjected to channel-by-channel semantic feature interaction and feature activation based on a fully connected layer to obtain a channel semantic association feature vector between financial data of borrowers and lenders; the channel semantic association feature vector between financial data of borrowers and lenders is input into a gating unit to perform masking processing on the feature values ​​of each position in the feature vector of channel semantic association between financial data of borrowers and lenders and then perform normalization operation to obtain a normalized semantic weight vector between financial data of borrowers and lenders; the feature values ​​of each position in the normalized semantic weight vector between financial data of borrowers and lenders are used as weighting coefficients to perform weighted optimization on each feature matrix along the channel dimension in the semantic association enhanced feature map between financial data of borrowers and lenders to obtain the semantic association enhanced feature map between financial data of borrowers and lenders; The semantic association feature map between the financial data of the borrowers and lenders is input into the category-related attention allocation network based on the gating mechanism and processed by the following category-related attention enhancement formula to obtain the semantic association enhancement feature map between the financial data of the borrowers and lenders; wherein the category-related attention enhancement formula is: ;in, is a semantic association feature graph between the borrower's financial data, , and They are global mean pooling, maximum value pooling and random value pooling. , and are preset trainable weight values, is the multi-scale pooling representation vector of semantic association features between borrowers’ financial data, and are the weight matrix and bias vector of the convolutional layer, respectively. for Activation function, is the hidden semantic association feature vector between the borrower’s financial data, is the hidden semantic association feature vector between the masked borrower and lender financial data, is the hidden semantic association feature vector between the masked borrower financial data. The eigenvalues ​​at the positions, is the semantic weight vector between normalized borrower and lender financial data, It means point multiplication by position. is a semantic association enhanced feature graph between the borrower's financial data; Wherein, determining whether the credit risk of the borrower exceeds a predetermined risk level based on the semantic association enhanced feature graph between the borrower's financial data includes: passing the semantic association enhanced feature graph between the borrower's financial data through a classifier-based early warning analyzer to obtain an early warning analysis result, wherein the early warning analysis result is used to indicate whether the credit risk of the borrower exceeds a predetermined risk level; Among them, inputting the semantic association enhancement feature map between the financial data of the lenders and the borrowers into the early warning analyzer based on the classifier to obtain the early warning analysis result specifically includes: determining the maximum eigenvalue of the semantic association enhancement between the financial data of the lenders and the minimum eigenvalue of the semantic association enhancement between the financial data of the lenders and the borrowers, and calculating the mean value of the semantic association enhancement between the financial data of the lenders and the standard deviation of the semantic association enhancement between the financial data of the feature set of the semantic association enhancement between the financial data of the lenders and the borrowers; calculating the quotient of the mean value of the semantic association enhancement between the financial data of the lenders and the standard deviation of the semantic association enhancement between the financial data of the lenders and the borrowers to obtain the statistical normalization value of the semantic association enhancement between the financial data of the lenders; calculating the inverse of each eigenvalue of the semantic association enhancement feature map between the financial data of the lenders and the borrowers, and comparing it with the maximum value of the semantic association enhancement between the financial data of the lenders and the borrowers. After multiplying the difference between the eigenvalue and the minimum eigenvalue of the semantic association enhancement between the financial data of the borrowers and the lenders, the eigenvalue is subtracted from the statistical normalized value of the semantic association enhancement between the financial data of the borrowers and the lenders to obtain a distribution approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders; an exponential function of each eigenvalue of the distribution approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders is calculated with a natural constant as the base, and the function is added to the statistical normalized value of the semantic association enhancement between the financial data of the borrowers and the lenders to obtain a probability approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders; the logarithm of the absolute value of each eigenvalue of the probability approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders is calculated with a base two to obtain an optimized feature map of the semantic association enhancement between the financial data of the borrowers and the lenders; and the optimized feature map of the semantic association enhancement between the financial data of the borrowers and the lenders is input into a warning analyzer based on a classifier to obtain a warning analysis result.

2. The early warning analysis method for financial data according to claim 1, characterized in that: The personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector are semantically co-space mapped to obtain the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector, including: the personal information semantic coding vector, the credit record semantic coding vector and the financial status semantic coding vector are respectively passed through a semantic co-space mapper based on a fully connected layer to obtain the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector.

3. The early warning analysis method for financial data according to claim 2, characterized in that: The personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector are subjected to financial data association pattern feature extraction to obtain a semantic association feature graph between the borrower's financial data, including: arranging the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector into a semantic association matrix and then passing the result through a financial data association pattern feature extractor based on a void convolutional neural network model to obtain the semantic association feature graph between the borrower's financial data.

4. The early warning analysis method for financial data according to claim 3, characterized in that: Calculating the position-weighted sum of the global maximum pooling feature vector of the semantic association feature between the financial data of the borrowers and lenders, the global mean pooling feature vector of the semantic association feature between the financial data of the borrowers and lenders, and the global random value pooling feature vector of the semantic association feature between the financial data of the borrowers and lenders to obtain a multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and lenders, including: calculating the position-weighted sum of the global maximum pooling feature vector of the semantic association feature between the financial data of the borrowers and lenders and a first preset trainable weight value to obtain a weight-modulated global maximum pooling feature vector of the semantic association feature between the financial data of the borrowers and lenders; calculating the position-weighted sum of the global mean pooling feature vector of the semantic association feature between the financial data of the borrowers and lenders and a second preset trainable weight value The product is used to obtain a global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation; the positional product between the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders and the third preset trainable weight value is calculated to obtain a global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation; the global maximum value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation, the global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders after weight modulation and the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders are added by position to obtain a multi-scale pooling representation vector of the semantic association features between the financial data of the borrowers and the lenders.

5. The early warning analysis method for financial data according to claim 4, characterized in that: The multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders is subjected to channel-by-channel semantic feature interaction and feature activation based on a fully connected layer to obtain the channel semantic association feature vector between the financial data of the borrowers and the lenders, including: calculating the matrix multiplication of the multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders and then adding the vector to the bias vector to obtain the bias-adjusted multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders; the bias-adjusted multi-scale pooling representation vector of the semantic association feature between the financial data of the borrowers and the lenders is obtained by Activation processing is performed in the activation function to obtain the channel semantic association feature vector between the borrower's financial data.

6. The early warning analysis method for financial data according to claim 5, characterized in that: Inputting the channel semantic association feature vector between the financial data of the borrowers and the lenders into the gating unit, masking each position feature value in the channel semantic association feature vector between the financial data of the borrowers and the lenders, and then performing a normalization operation to obtain a normalized semantic weight vector between the financial data of the borrowers and the lenders, including: masking each position feature value in the channel semantic association feature vector between the financial data of the borrowers and the lenders based on the gating mechanism to obtain a masked channel semantic association feature vector between the financial data of the borrowers and the lenders, wherein, in response to each position feature value in the channel semantic association feature vector between the financial data of the borrowers and the lenders being greater than or equal to a predetermined threshold, each position feature value in the channel semantic association feature vector between the financial data of the borrowers and the lenders takes an original value, otherwise it is set to zero; each position feature value in the channel semantic association feature vector between the financial data of the borrowers and the lenders is normalized to obtain a normalized semantic weight vector between the financial data of the borrowers and the lenders.

7. An early warning analysis system for financial data, characterized in that: include: A financial-related data acquisition module is used to acquire the financial-related data of the borrower, wherein the financial-related data includes personal information, credit records and financial status; a semantic encoding module is used to semantically encode the personal information, credit records and financial status in the financial-related data to obtain a personal information semantic encoding vector, a credit record semantic encoding vector and a financial status semantic encoding vector; a semantic co-space mapping module is used to semantically co-space map the personal information semantic encoding vector, the credit record semantic encoding vector and the financial status semantic encoding vector to obtain a personal information mapping semantic vector, a credit record mapping semantic encoding vector and a financial status mapping semantic encoding vector. vector; a financial data association pattern feature extraction module, used for extracting the financial data association pattern features from the personal information mapping semantic vector, the credit record mapping semantic coding vector and the financial status mapping semantic coding vector to obtain a semantic association feature map between the borrower's financial data; a category-related attention allocation module, used for inputting the semantic association feature map between the borrower's financial data into a category-related attention allocation network based on a gating mechanism to obtain a semantic association enhanced feature map between the borrower's financial data; a credit risk risk level determination module, used for determining whether the borrower's credit risk exceeds a predetermined risk level based on the semantic association enhanced feature map between the borrower's financial data; The method comprises inputting the semantic association feature map between the financial data of the borrowers and the lenders into a category-dependent attention allocation network based on a gating mechanism to obtain an enhanced semantic association feature map between the financial data of the borrowers and the lenders, including: performing a maximum-based global pooling process, a random-value-based global pooling process and an average-based global pooling process on the semantic association feature map between the financial data of the borrowers and the lenders to obtain a global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, a global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, and a global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders; and calculating the position-based sum of the global maximum pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, the global mean pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders, and the global random value pooling feature vector of the semantic association features between the financial data of the borrowers and the lenders. The weights are summed to obtain a multi-scale pooling representation vector of semantic association features between financial data of borrowers and lenders; the multi-scale pooling representation vector of semantic association features between financial data of borrowers and lenders is subjected to channel-by-channel semantic feature interaction and feature activation based on a fully connected layer to obtain a channel semantic association feature vector between financial data of borrowers and lenders; the channel semantic association feature vector between financial data of borrowers and lenders is input into a gating unit to perform masking processing on the feature values ​​of each position in the feature vector of channel semantic association between financial data of borrowers and lenders and then perform normalization operation to obtain a normalized semantic weight vector between financial data of borrowers and lenders; the feature values ​​of each position in the normalized semantic weight vector between financial data of borrowers and lenders are used as weighting coefficients to perform weighted optimization on each feature matrix along the channel dimension in the semantic association enhanced feature map between financial data of borrowers and lenders to obtain the semantic association enhanced feature map between financial data of borrowers and lenders; The semantic association feature map between the financial data of the borrowers and lenders is input into the category-related attention allocation network based on the gating mechanism and processed by the following category-related attention enhancement formula to obtain the semantic association enhancement feature map between the financial data of the borrowers and lenders; wherein the category-related attention enhancement formula is: ;in, is a semantic association feature graph between the borrower's financial data, , and They are global mean pooling, maximum value pooling and random value pooling. , and are preset trainable weight values, is the multi-scale pooling representation vector of semantic association features between borrowers’ financial data, and are the weight matrix and bias vector of the convolutional layer, respectively. for Activation function, is the hidden semantic association feature vector between the borrower’s financial data, is the hidden semantic association feature vector between the masked borrower and lender financial data, is the hidden semantic association feature vector between the masked borrower financial data. The eigenvalues ​​at the positions, is the semantic weight vector between normalized borrower and lender financial data, It means point multiplication by position. is a semantic association enhanced feature graph between the borrower's financial data; Wherein, determining whether the credit risk of the borrower exceeds a predetermined risk level based on the semantic association enhanced feature graph between the borrower's financial data includes: passing the semantic association enhanced feature graph between the borrower's financial data through a classifier-based early warning analyzer to obtain an early warning analysis result, wherein the early warning analysis result is used to indicate whether the credit risk of the borrower exceeds a predetermined risk level; Among them, inputting the semantic association enhancement feature map between the financial data of the lenders and the borrowers into the early warning analyzer based on the classifier to obtain the early warning analysis result specifically includes: determining the maximum eigenvalue of the semantic association enhancement between the financial data of the lenders and the minimum eigenvalue of the semantic association enhancement between the financial data of the lenders and the borrowers, and calculating the mean value of the semantic association enhancement between the financial data of the lenders and the standard deviation of the semantic association enhancement between the financial data of the feature set of the semantic association enhancement between the financial data of the lenders and the borrowers; calculating the quotient of the mean value of the semantic association enhancement between the financial data of the lenders and the standard deviation of the semantic association enhancement between the financial data of the lenders and the borrowers to obtain the statistical normalization value of the semantic association enhancement between the financial data of the lenders; calculating the inverse of each eigenvalue of the semantic association enhancement feature map between the financial data of the lenders and the borrowers, and comparing it with the maximum value of the semantic association enhancement between the financial data of the lenders and the borrowers. After multiplying the difference between the eigenvalue and the minimum eigenvalue of the semantic association enhancement between the financial data of the borrowers and the lenders, the eigenvalue is subtracted from the statistical normalized value of the semantic association enhancement between the financial data of the borrowers and the lenders to obtain a distribution approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders; an exponential function of each eigenvalue of the distribution approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders is calculated with a natural constant as the base, and the function is added to the statistical normalized value of the semantic association enhancement between the financial data of the borrowers and the lenders to obtain a probability approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders; the logarithm of the absolute value of each eigenvalue of the probability approximation map of the semantic association enhancement between the financial data of the borrowers and the lenders is calculated with a base two to obtain an optimized feature map of the semantic association enhancement between the financial data of the borrowers and the lenders; and the optimized feature map of the semantic association enhancement between the financial data of the borrowers and the lenders is input into a warning analyzer based on a classifier to obtain a warning analysis result.

8. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processing device, the steps of the early warning analysis method for financial data described in any one of claims 1 to 6 are implemented.

Images