Method and system for constructing social credit index library based on big data

Abstract

The application relates to the technical field of data processing, and particularly discloses a social credit index library construction method and system based on big data, which extracts a three-level index data set from a social credit index library, and uses a deep learning algorithm to perform semantic embedding coding on each three-level index to understand the semantic information of each three-level index; then, after determining a three-level index to be verified, the three-level index to be verified is removed from the data set, and a three-level index semantic environment is constructed based on the remaining three-level index data set; through dynamic semantic query response analysis of the three-level index to be verified in the constructed three-level index semantic environment, it is identified whether the three-level index to be verified has semantic intersection or semantic contradiction, so as to realize deep verification of the three-level index. In this way, efficient and accurate verification of the three-level index can be realized, which is helpful to construct a more scientific and accurate social credit index library and improve the fairness and reliability of social credit evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and more specifically, to a social credit index library construction method and system based on big data. BACKGROUND

[0002] With the rapid development of social economy, the social credit system plays an increasingly key role in maintaining market order, promoting the healthy operation of the economy, and ensuring social fairness and justice. As a core component of the social credit system, the scientificity, accuracy, and completeness of the construction of the social credit index library directly affect the reliability and effectiveness of social credit evaluation.

[0003] The third-level indicators are the most specific measurement units in the social credit evaluation system. They are further refined based on higher-level (first-level and second-level) indicators, providing specific measurement standards for specific behaviors, activities, or attributes, and directly reflecting the performance of individuals or enterprises in specific aspects. For example, in enterprise credit evaluation, if the first-level indicator is "financial health status", the second-level indicators may be further divided into "debt servicing ability", "profitability", and "operational efficiency", and the third-level indicators under the "debt servicing ability" indicator may include but are not limited to: current ratio, quick ratio, and asset-liability ratio. In personal credit evaluation, assuming the first-level indicator is "credit history", the second-level indicators may be "credit card repayment record" and "loan repayment record", and the third-level indicators under "credit card repayment record" may include: recent overdue records, overdue days, and average monthly credit card usage limit, etc.

[0004] The third-level indicators in the social credit index library are key elements for refining the measurement of credit status, and their semantic accuracy directly affects the fairness and reliability of the entire credit evaluation. However, in actual operation, the setting of third-level indicators faces multiple challenges. On the one hand, social credit involves a wide range of fields, including economic activities, law-abiding, social responsibility, and many other aspects, each of which can be further divided into numerous specific indicators, resulting in a large number of third-level indicators with diverse content. On the other hand, due to differences in professional background, understanding perspective, and information sources among indicator developers, there may be semantic overlap or potential conflicts between different indicators, which not only reduces the accuracy and efficiency of credit evaluation, but also may cause social misunderstanding and trust crisis.

[0005] Traditional third-level indicator verification methods mainly rely on manual review, but manual review is easily influenced by subjective factors and is difficult to comprehensively identify all potential semantic overlaps and conflicts, especially when the indicators are similar in expression or involve professional terms, making it more difficult to make accurate judgments, thereby increasing the uncertainty of verification. Secondly, in the face of massive data and high complexity of the indicator system, manual review is inefficient and difficult to meet the needs of rapid construction and updating of the social credit system.

[0006] Therefore, a social credit index library construction method and system based on big data are expected. SUMMARY

[0007] To solve the above technical problems, the present application is proposed. Embodiments of the present application provide a social credit index library construction method and system based on big data, which extracts a three-level index data set from a social credit index library, and uses a deep learning algorithm to perform semantic embedding coding on each three-level index to understand the semantic information of each three-level index. Then, after determining a three-level index to be verified, the three-level index to be verified is removed from the data set, and a three-level index semantic environment is constructed based on the remaining three-level index data set. By performing dynamic semantic query response analysis on the three-level index to be verified in the constructed three-level index semantic environment, it is determined whether the three-level index to be verified has semantic intersection or semantic contradiction, thereby realizing deep verification of the three-level index. In this way, efficient and accurate verification of the three-level index can be realized, which helps to construct a more scientific and accurate social credit index library and improve the fairness and reliability of social credit evaluation.

[0008] According to one aspect of the present application, a social credit index library construction method based on big data is provided, which includes:

[0009] extracting a three-level index from a social credit index library to obtain a data subset of the three-level index;

[0010] performing semantic embedding coding on each three-level index in the data subset of the three-level index to obtain a data subset of three-level index semantic embedding coding vectors;

[0011] extracting a three-level index semantic embedding coding vector of a three-level index to be verified from the data subset of the three-level index semantic embedding coding vectors as a three-level index semantic embedding coding vector to be verified;

[0012] based on the data subset of the three-level index semantic embedding coding vectors, performing semantic query deep verification on the three-level index semantic embedding coding vector to be verified to obtain an index verification result, wherein the semantic query deep verification on the three-level index semantic embedding coding vector to be verified includes: after deleting the three-level index semantic embedding coding vector to be verified from the data subset of the three-level index semantic embedding coding vectors, performing semantic query dynamic response analysis with the three-level index semantic embedding coding vector to be verified to obtain the index verification result.

[0013] According to another aspect of the present application, a social credit index library construction system based on big data is provided, which includes:

[0014] The third-level index extraction module is configured to extract third-level indexes from the social credit index library to obtain a data subset of third-level indexes.

[0015] The semantic embedding coding module is configured to perform semantic embedding coding on each third-level index in the data subset of third-level indexes to obtain a data subset of third-level index semantic embedding coding vectors.

[0016] The to-be-verified index extraction module is configured to extract a third-level index semantic embedding coding vector of a third-level index to be verified from the data subset of third-level index semantic embedding coding vectors as a to-be-verified third-level index semantic embedding coding vector.

[0017] The semantic query depth verification module is configured to perform semantic query depth verification on the to-be-verified third-level index semantic embedding coding vector based on the data subset of third-level index semantic embedding coding vectors to obtain an index verification result. The semantic query depth verification on the to-be-verified third-level index semantic embedding coding vector includes performing semantic query dynamic response analysis on the to-be-verified third-level index semantic embedding coding vector after deleting the to-be-verified third-level index semantic embedding coding vector from the data subset of third-level index semantic embedding coding vectors to obtain the index verification result.

[0018] Compared with the prior art, the social credit index library construction method and system based on big data provided by the present application can extract a third-level index data set from a social credit index library, and use a deep learning algorithm to perform semantic embedding coding on each third-level index to understand the semantic information of each third-level index. Then, after determining a third-level index to be verified, the third-level index to be verified is removed from the data set, and a third-level index semantic environment is constructed based on the remaining third-level index data set. By performing dynamic semantic query response analysis on the third-level index to be verified in the constructed third-level index semantic environment, it is determined whether the third-level index to be verified has semantic intersection or semantic contradiction, thereby realizing deep verification of the third-level index. In this way, efficient and accurate verification of the third-level index can be realized, which helps to construct a more scientific and accurate social credit index library and improve the fairness and reliability of social credit evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0019] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description of embodiments of the present application taken in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of embodiments of the present application and constitute a part of the specification, together with the description of embodiments of the present application, to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1Flow chart of the method for constructing a big data-based social credit index library according to an embodiment of the present application.

[0021] Figure 2 Data flow diagram of the method for constructing a big data-based social credit index library according to an embodiment of the present application.

[0022] Figure 3 Flow chart of sub-step S4 of the method for constructing a big data-based social credit index library according to an embodiment of the present application.

[0023] Figure 4 Flow chart of sub-step S42 of the method for constructing a big data-based social credit index library according to an embodiment of the present application.

[0024] Figure 5 Flow chart of sub-step S423 of the method for constructing a big data-based social credit index library according to an embodiment of the present application.

[0025] Figure 6 Block diagram of the system for constructing a big data-based social credit index library according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] As shown in the present application and claims, unless otherwise clearly indicated by context, "one," "an," "a" and / or "the" are not to be construed as specific reference unless otherwise clear from context. Generally, the terms "include" and "comprise" are used to indicate included elements without precluding other elements not specifically named. Generally, the term "comprise" is used to indicate included elements without precluding other elements not specifically named.

[0027] Although the present application makes various references to certain modules in the system according to embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server. The modules are merely illustrative, and different aspects of the system and method can use different modules.

[0028] Flow charts are used in the present application to illustrate the operations performed by the system according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in sequence. Instead, various steps can be processed in reverse order or simultaneously, as desired. Other operations can also be added to or removed from these processes.

[0029] In the following, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. It is obvious that the described embodiments are only a part of the embodiments of the present application, and are not all the embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.

[0030] It is worth noting that in this application, all actions of obtaining data are carried out in accordance with the corresponding data protection regulations and policies of the country where the device is located, and with the authorization given by the owner of the corresponding device.

[0031] To solve the technical problems in the background art, the present application proposes a social credit index library construction method based on big data, which extracts a three-level index data set from a social credit index library, and uses a deep learning algorithm to perform semantic embedding coding on each three-level index to understand the semantic information of each three-level index. Then, after determining the three-level index to be verified, the three-level index to be verified is removed from the data set, and a three-level index semantic environment is constructed based on the remaining three-level index data set. By performing dynamic semantic query response analysis on the three-level index to be verified in the constructed three-level index semantic environment, it is determined whether the three-level index to be verified has semantic intersection or semantic contradiction, thereby realizing deep verification of the three-level index. In this way, efficient and accurate verification of the three-level index can be realized, which helps to construct a more scientific and accurate social credit index library and improve the fairness and reliability of social credit evaluation.

[0032] Figure 1 A flowchart of the social credit index library construction method based on big data according to an embodiment of the present application. Figure 2 A data flow diagram of the social credit index library construction method based on big data according to an embodiment of the present application. As shown in Figure 1 and Figure 2 The social credit index library construction method based on big data includes the following steps: S1, extracting a three-level index from a social credit index library to obtain a data subset of the three-level index; S2, performing semantic embedding coding on each three-level index in the data subset of the three-level index to obtain a data subset of three-level index semantic embedding coding vectors; S3, extracting a three-level index semantic embedding coding vector of a three-level index to be verified from the data subset of the three-level index semantic embedding coding vector as a three-level index semantic embedding coding vector to be verified; S4, based on the data subset of the three-level index semantic embedding coding vector, performing semantic query deep verification on the three-level index semantic embedding coding vector to be verified to obtain an index verification result, wherein the semantic query deep verification on the three-level index semantic embedding coding vector to be verified includes: after deleting the three-level index semantic embedding coding vector to be verified from the data subset of the three-level index semantic embedding coding vector, performing semantic query dynamic response analysis with the three-level index semantic embedding coding vector to be verified to obtain the index verification result.

[0033] In the above method for constructing a social credit index library based on big data, step S1 extracts three-level indicators from the social credit index library to obtain a data subset of the three-level indicators. It should be understood that the social credit index library is a large and hierarchical system that covers credit measurement indicators from macro to micro and multiple dimensions. As the basic unit of the credit evaluation system, three-level indicators play a fundamental role in accurately depicting the social credit situation. Therefore, the present application expects to understand the composition of social credit from the finest granularity by deeply analyzing the three-level indicators in the social credit index library, thereby laying a solid foundation for constructing an accurate and reliable credit evaluation mechanism.

[0034] Specifically, in the initial stage, a comprehensive and systematic review of the entire social credit index library is needed. This is not just a simple browse, but a deep understanding of the meaning behind each indicator, the scope of application, and how it affects the final social credit evaluation results. For example, in the field of enterprise credit evaluation, for the first-level indicator "financial health status", all related secondary indicators such as "debt servicing ability", "profitability" and "operational efficiency" must be carefully analyzed. In particular, for the secondary indicator "debt servicing ability", further explore multiple tertiary indicators including current ratio, quick ratio and asset-liability ratio, understand their respective definitions, calculation methods and their impact on enterprise credit status.

[0035] Next, a detailed work framework needs to be constructed to guide the entire extraction process. This framework should not only include specific classification criteria, but also clearly define the specific content and requirements of each category. For example, all three-level indicators related to enterprise debt servicing ability can be classified into one category, and then sorted according to their importance or application frequency. Such an organizational approach helps subsequent data processing and analysis work more efficient and orderly. At the same time, similar situations in personal credit evaluation also need to be considered, such as when the first-level indicator is "credit history", how to systematically organize specific three-level indicators under the secondary indicators such as credit card repayment records, loan repayment records, etc., such as recent overdue records, overdue days, average monthly credit card usage limit, etc., to facilitate quick positioning and use.

[0036] Data verification is a key step to ensure the accuracy of the selected three-level indicators. This means that evidence needs to be collected from multiple information sources to confirm the authenticity and reliability of these indicators. For example, when confirming the current ratio of an enterprise, it may be necessary to review the company's annual financial reports, data analysis provided by third-party audit institutions, and other materials. In addition, attention should be paid to the differences between different industries and regions. Some specific three-level indicators may be very important in one industry or region, but not very suitable in another environment. Therefore, it is particularly important to adjust the selection criteria of the indicators flexibly to better adapt to different evaluation needs.

[0037] To ensure transparency and traceability in the work, it is crucial to establish a detailed documentation mechanism. From the initial selection of indicators to the final data processing, each step should be well-documented. These documents should not only include the name, definition, and source of each tertiary indicator, but also the reason for selecting the indicator and any relevant notes. This not only helps team members better communicate and collaborate, but also provides reference materials for future revisions and improvements. For example, if a tertiary indicator is found to be no longer applicable to the current socio-economic environment in the future, relevant information can be quickly found based on previous records and adjustments can be made accordingly.

[0038] In addition, attention should be paid to the logical relationship and consistency between indicators. In a large social credit indicator library, there may be complex interaction relationships between various tertiary indicators. By establishing models or charts, these relationships can be visually displayed to help better understand and analyze. For example, the relationship between a company's current ratio and quick ratio can directly reflect the company's ability to repay debts in the short term, and this relationship is important for a comprehensive assessment of the company's financial health. Therefore, during the extraction of tertiary indicators, special attention should be paid to these internal relationships to ensure that the selected indicators are both independent and complementary, forming a complete evaluation system.

[0039] Finally, considering the possibility of future changes and trends, the socio-economic environment is constantly changing, and new risk factors and opportunities may appear at any time, which requires the tertiary indicator data subset to be updated and adjusted in a timely manner to maintain its effectiveness and advancement. To this end, a regular review mechanism can be established on the existing basis, and based on the latest research results and social needs, some tertiary indicators can be added or removed in a timely manner to ensure that the entire credit evaluation system is always in the best state.

[0040] In the method for constructing the social credit index library based on big data, the step S2 is to perform semantic embedding coding on each of the three-level indexes in the data subset of the three-level indexes to obtain a data subset of three-level index semantic embedding coding vectors. In a specific example of the present application, the step S2 includes inputting each of the three-level indexes in the data subset of the three-level indexes into a semantic embedding encoder based on a Bert model to obtain the data subset of three-level index semantic embedding coding vectors. Specifically, since the original three-level indexes are in the form of text, in order to enable a computer to accurately understand the semantic information of the text, it is necessary to convert it into a numerical vector form that can be processed by a computer. For this purpose, the present application uses a Bert model as a semantic embedding encoder to perform semantic embedding coding on each of the three-level indexes in the data subset of the three-level indexes to obtain the data subset of three-level index semantic embedding coding vectors. Those skilled in the art should know that the Bert model, as a leading technology in the field of natural language processing, has learned rich language knowledge and semantic representation through large-scale unsupervised pre-training, and has strong semantic understanding ability, which can deeply mine the semantic features in the text. By using the Bert model to perform semantic embedding coding on each of the three-level indexes, the present application can convert the text semantics of each of the three-level indexes into a numerical representation in a high-dimensional vector space, and measure the semantic similarity between different three-level indexes in terms of distance distribution in the vector space, so as to reveal the subtle semantic differences between the three-level indexes and provide effective information representation for subsequent verification work.

[0041] In the method for constructing the social credit index library based on big data, the step S3 is to extract the three-level index semantic embedding coding vector of the three-level index to be verified from the data subset of three-level index semantic embedding coding vectors as a three-level index semantic embedding coding vector to be verified. It should be understood that, since the number of indexes in the data subset of the three-level indexes is large, in order to improve the efficiency and accuracy of verification, the present application first determines the target object to be verified, and extracts the three-level index semantic embedding coding vector of the three-level index to be verified as a three-level index semantic embedding coding vector to be verified, so as to verify the three-level indexes in the social credit index library one by one, thereby improving the pertinence and effectiveness of verification analysis.

[0042] In the above-mentioned big data-based social credit index library construction method, the step S4, based on the data subset of the three-level index semantic embedding coding vector, the semantic query deep checking of the to-be-verified three-level index semantic embedding coding vector is performed to obtain an index verification result, wherein the semantic query deep checking of the to-be-verified three-level index semantic embedding coding vector comprises: after deleting the to-be-verified three-level index semantic embedding coding vector from the data subset of the three-level index semantic embedding coding vector, the semantic query dynamic response analysis of the to-be-verified three-level index semantic embedding coding vector is performed to obtain the index verification result. Wherein, Figure 3 The flow chart of the sub-step S4 of the big data-based social credit index library construction method according to the embodiment of the present application is shown in FIG. 4. Figure 3 As shown in FIG. 4, the step S4 comprises steps: S41, deleting the to-be-verified three-level index semantic embedding coding vector in the data subset of the three-level index semantic embedding coding vector to obtain a data set of three-level index library semantic embedding coding vector; S42, performing three-level index semantic query checking coding based on adaptive weight anchoring on the to-be-verified three-level index semantic embedding coding vector and the data set of the three-level index library semantic embedding coding vector to obtain a to-be-verified three-level index semantic query deep checking coding vector; S43, based on the to-be-verified three-level index semantic query deep checking coding vector, determining the index verification result, the index verification result is used to represent whether the to-be-verified three-level index exists semantic intersection or semantic contradiction.

[0043] Specifically, the step S41, the to-be-verified three-level index semantic embedding coding vector in the data subset of the three-level index semantic embedding coding vector is deleted to obtain a data set of three-level index library semantic embedding coding vector. Specifically, after determining the to-be-verified three-level index, in order to accurately identify whether it exists semantic intersection and contradiction with other three-level indexes, the to-be-verified three-level index semantic embedding coding vector is further deleted from the data subset of the three-level index semantic embedding coding vector, so as to avoid the interference of the semantic information of the to-be-verified three-level index itself on the semantic interaction analysis between indexes. Then, the remaining three-level index semantic embedding coding vectors are combined into a data set of three-level index library semantic embedding coding vectors, so as to construct a pure semantic environment for the semantic query response analysis of the to-be-verified three-level index, so as to improve the accuracy of semantic intersection and contradiction identification.

[0044] Specifically, the step S42, the data set of the to-be-verified three-level index semantic embedding encoding vector and the three-level index library semantic embedding encoding vector is subjected to three-level index semantic query verification coding based on adaptive weight anchoring to obtain a to-be-verified three-level index semantic query deep verification encoding vector. Specifically, in order to deeply mine the semantic association and potential contradiction between the to-be-verified three-level index and the remaining three-level indexes, the present application proposes a three-level index semantic query verification coding method based on adaptive weight anchoring, which captures the semantic association features between the to-be-verified three-level index and each three-level index by respectively performing semantic query interaction coding on the to-be-verified three-level index semantic embedding encoding vector and each three-level index library semantic embedding encoding vector in the data set, and on this basis, dynamically adjusts the weights of different semantic association features by introducing an adaptive weight anchoring mechanism, to strengthen the attention to the semantic relationship between the to-be-verified three-level index and the key three-level index, thereby realizing accurate identification of potential semantic cross and contradiction. Wherein, Figure 4 The flow chart of the sub-step S42 of the method for constructing a social credit index library based on big data according to the embodiment of the present application is shown in FIG. 4. Figure 4 As shown in FIG. 4, the step S42 includes the following steps: S421, performing deep implicit feature extraction on each three-level index library semantic embedding encoding vector in the data set of the to-be-verified three-level index semantic embedding encoding vector and the three-level index library semantic embedding encoding vector to obtain a data set of to-be-verified three-level index semantic query feature deep implicit encoding vectors and three-level index library semantic feature deep implicit encoding vectors; S422, performing semantic response decision anchoring coding on each three-level index library semantic feature deep implicit encoding vector in the data set of the to-be-verified three-level index semantic query feature deep implicit encoding vector and the three-level index library semantic feature deep implicit encoding vector to obtain a data set of to-be-verified three-level index-index library semantic response anchoring encoding matrices; S423, performing dynamic aggregation coding on the data set of the to-be-verified three-level index-index library semantic response anchoring encoding matrices to obtain a to-be-verified three-level index semantic query dynamic response encoding vector.

[0045] More specifically, in one specific example of the present application, the step S421 includes using a deep implicit feature extraction module based on a fully connected coding network to process each three-level index library semantic embedding encoding vector in the data set of the to-be-verified three-level index semantic embedding encoding vector and the three-level index library semantic embedding encoding vector to obtain a data set of to-be-verified three-level index semantic query feature deep implicit encoding vectors and three-level index library semantic feature deep implicit encoding vectors, which can be expressed by the formula:

[0046] V2={v 21 ,v 22 ,...,v 2i ,...,v2n}

[0047]

[0048] wherein V2 represents a dataset of tertiary index library semantic embedding encoding vectors, v 21 , v 22 , v 2i and v 2n represent the 1st, 2nd, i-th and n-th tertiary index library semantic embedding encoding vectors in the dataset of tertiary index library semantic embedding encoding vectors respectively, n is the number of vectors in the dataset of tertiary index library semantic embedding encoding vectors, W1 represents a to-be-verified tertiary index semantic query information semantic feature weight matrix, W2 represents a tertiary index library semantic feature weight matrix, b1 represents a to-be-verified tertiary index semantic query information semantic feature bias term, b2 represents a tertiary index library semantic feature bias term, V1 represents a to-be-verified tertiary index semantic embedding encoding vector, sigmoid(·) represents a sigmoid activation function, V d1 represents a to-be-verified tertiary index semantic query feature deep implicit encoding vector, v d2i represents v 2i corresponding tertiary index library semantic feature deep implicit encoding vector.

[0049] That is, by performing deep implicit feature extraction based on full connection encoding on the to-be-verified tertiary index semantic embedding encoding vector and each tertiary index library semantic embedding encoding vector respectively, deeper-level semantic features of each tertiary index are captured through nonlinear mapping of the full connection neural network, the processing capability of the model for complex semantic information is enhanced, and the analysis accuracy of the semantic relationship between indexes is further improved.

[0050] More specifically, the step S422 is expressed by a formula as follows:

[0051]

[0052] wherein (·) T represents the transpose of a vector, represents vector multiplication, S represents a feature scale scaling factor, M 12i represents the to-be-verified tertiary index-index library semantic response anchor encoding matrix between V d1 and v d2i .

[0053] That is, by further performing semantic response interaction encoding on the semantic query feature deep implicit coding vector of the third-level indicator to be verified and the semantic feature deep implicit coding vector of each third-level indicator library, the semantic association interaction information between the third-level indicator to be verified and the remaining third-level indicators in the data set is captured to generate the semantic response anchor coding matrix of the third-level indicator to be verified-indicator library.

[0054] Figure 5 Flowchart of sub-step S423 of the method for constructing a social credit index database based on big data according to an embodiment of the present application. Figure 5 As shown, the step S423 includes the steps of: S4231, based on the feature distribution of each three-level indicator-index library semantic response anchor coding matrix to be verified in the data set of the three-level indicator-index library semantic response anchor coding matrix to be verified, calculating the decision anchor adaptive splicing factor of each three-level indicator-index library semantic response anchor coding matrix to be verified to obtain a data set of the three-level indicator-index library decision anchor adaptive splicing factor to be verified, wherein the three-level indicator-index library decision anchor adaptive splicing factor to be verified and the maximum eigenvalue of the three-level indicator-index library semantic response anchor coding matrix to be verified, The feature variance, feature mean and number of eigenvalues ​​are related; S4232, the data set of the adaptive splicing factor of the decision anchor of the three-level indicator to be verified is weighted based on the Softmax function to obtain the data set of the adaptive splicing weight factor of the decision anchor of the three-level indicator to be verified; S4233, based on the data set of the adaptive splicing weight factor of the decision anchor of the three-level indicator to be verified, the data set of the semantic response anchor coding matrix of the three-level indicator to be verified is weighted fused and feature reshaped to obtain the dynamic response coding vector of the semantic query of the three-level indicator to be verified.

[0055] In a specific example of the present application, step S4231 includes: taking the sum of the characteristic variance and the drift coefficient of the semantic response anchor coding matrix of the three-level indicator-indicator library to be verified as the numerator, and calculating the square of the difference between the maximum eigenvalue of the semantic response anchor coding matrix of the three-level indicator-indicator library to be verified and its characteristic mean multiplied by the number of its eigenvalues, and adding the drift coefficient and twice the characteristic variance as the denominator to obtain the adaptive splicing factor of the decision anchor of the three-level indicator-indicator library to be verified, which is expressed as follows:

[0056] k=count(M 12i )

[0057]

[0058] wherein count(·) represents the number of elements of a matrix, k represents a difference amplification coefficient, i.e., the number of eigenvalues of the to-be-verified three-level index-index library semantic response anchor encoding matrix, σ 2 represents the characteristic variance of the to-be-verified three-level index-index library semantic response anchor encoding matrix, ∈ represents a drift coefficient of the to-be-verified three-level index-index library semantic response anchor encoding matrix, μ represents a characteristic mean value of the to-be-verified three-level index-index library semantic response anchor encoding matrix, max(·) is a maximum value function, E 12i represents M 12i corresponds to a to-be-verified three-level index-index library decision anchor adaptive splicing factor.

[0059] That is, by performing feature distribution analysis on the to-be-verified three-level index-index library semantic response anchor encoding matrix, the semantic response correlation strength between the to-be-verified three-level index and the rest of the three-level indexes in the index library is measured, and then the dominant position of the semantic interaction information contained in the to-be-verified three-level index-index library semantic response anchor encoding matrix in the overall semantic cross-evaluation task is determined, which is used as the weight basis in subsequent feature fusion.

[0060] In particular, the drift coefficient is used to smooth the feature distribution fluctuation of the to-be-verified three-level index-index library semantic response anchor encoding matrix. In one preferred example of the present application, for the drift coefficient ∈ in the to-be-verified three-level index-index library decision anchor adaptive splicing factor, the input feature set distribution of the to-be-verified three-level index-index library semantic response anchor encoding matrix transitions from a weakly interpretable overall integrity to a strongly interpretable local maximum, and the global dominance basis of the to-be-verified three-level index-index library semantic response anchor encoding matrix is enhanced through weak-to-strong interpretable generalization of the drift coefficient ∈, which is expressed by the formula as follows:

[0061]

[0062]

[0063] wherein η is an intermediate transition representation value of the feature distribution balance state of the to-be-verified three-level index-index library semantic response anchor encoding matrix, m ij is the jth eigenvalue of the matrix M 12i , and e is a natural constant.

[0064] Here, η is taken as an intermediate state transition representation from weakly interpretable to strongly interpretable, and for each eigenvalue m ijThe intermediate state transition η is globally controlled in importance score weight relative to the global state transition, taking the importance score of the to-be-verified three-level indicator-indicator library semantic response anchor coding matrix for the global smooth state transition as the importance score of the to-be-verified three-level indicator-indicator library semantic response anchor coding matrix, to realize the weight dependence of the to-be-verified three-level indicator-indicator library decision anchor adaptive splicing factor in the explainable generalization inference.

[0065] In one specific example of the present application, the step S4232 is expressed by the formula:

[0066] a 12i = softmax(E 12i )

[0067] wherein softmax(·) represents a normalized exponential function, a 12i represents the to-be-verified three-level indicator-indicator library decision anchor adaptive splicing weight factor of the matrix M 12i .

[0068] That is, the Softmax function is used to normalize the data set of the to-be-verified three-level indicator-indicator library decision anchor adaptive splicing factor into a weight data set with probability distribution properties. It should be understood that the Softmax transformation not only converts the to-be-verified three-level indicator-indicator library decision anchor adaptive splicing factor into an easily interpretable weight distribution, but also further amplifies the significant differences between the to-be-verified three-level indicator-indicator library semantic response anchor coding matrices through the characteristics of the exponential function, to enhance the distribution distinguishing ability of the features.

[0069] In one specific example of the present application, the step S4233 is expressed by the formula:

[0070]

[0071] wherein M c represents the to-be-verified three-level indicator-indicator library semantic response anchor coding fusion matrix, reshape(·) represents a feature shape reshaping function, and v c represents the to-be-verified three-level indicator semantic query dynamic response coding vector.

[0072] That is, the data set of the to-be-verified three-level indicator-indicator library semantic response anchor coding matrix is weighted and fused based on the generated weight distribution, to fuse the semantic interaction information of the to-be-verified three-level indicator relative to different indicators, and to restore it to a vector form through feature shape reshaping, to generate the to-be-verified three-level indicator semantic query dynamic response coding vector. Through this processing, the to-be-verified three-level indicator semantic query dynamic response coding vector comprehensively integrates the semantic interaction information of the to-be-verified three-level indicator relative to the remaining three-level indicators in the indicator library, providing a more abundant and accurate semantic basis for subsequent feature fusion and decision making.

[0073] Specifically, in one specific example of the present application, the step S43 comprises: inputting the to-be-verified three-level indicator semantic query depth verification encoding vector into the classifier-based indicator verifier to obtain the indicator verification result. Specifically, the classifier is based on a neural network architecture, which, after receiving the to-be-verified three-level indicator semantic query depth verification encoding vector, performs deep learning and feature extraction through nonlinear transformation of multiple hidden layers, thereby making a classification decision based on the semantic association information between the to-be-verified three-level indicator and other three-level indicators contained in the to-be-verified three-level indicator semantic query depth verification encoding vector, and outputting the indicator verification result of the to-be-verified three-level indicator. If the indicator verification result indicates that the to-be-verified three-level indicator has semantic intersection or semantic contradiction, it is marked as an abnormal indicator and fed back to the user for further review and processing. In this way, the present application can realize efficient and accurate verification of three-level indicators in the social credit indicator library, ensuring that each three-level indicator accurately reflects its intended credit dimension, thereby further improving the accuracy and reliability of the social credit evaluation system.

[0074] More specifically, inputting the to-be-verified three-level indicator semantic query depth verification encoding vector into the classifier-based indicator verifier to obtain the indicator verification result comprises: using the fully connected layer of the indicator verifier to perform fully connected encoding on the to-be-verified three-level indicator semantic query depth verification encoding vector to obtain a to-be-verified three-level indicator semantic query depth verification fully connected encoding vector; inputting the to-be-verified three-level indicator semantic query depth verification fully connected encoding vector into the Softmax classification function of the indicator verifier to obtain probability values of the to-be-verified three-level indicator semantic query depth verification encoding vector belonging to each classification label, wherein the classification labels include the to-be-verified three-level indicator having semantic intersection or semantic contradiction and not having semantic intersection or semantic contradiction; and determining the classification label corresponding to the largest probability value in the probability values as the indicator verification result.

[0075] Specifically, if the obtained indicator verification result indicates that the to-be-verified three-level indicator does not have semantic intersection or semantic contradiction, a comprehensive and detailed review of the existing three-level indicator library is needed to verify whether the definition, measurement standard and application scenario of each indicator are clear, unambiguous and without ambiguity. In this process, the specific content of each three-level indicator needs to be analyzed in depth to ensure that it can independently and effectively reflect the performance of individuals or enterprises in a specific aspect. For example, in enterprise credit evaluation, indicators such as liquidity ratio and quick ratio under the “debt paying ability” need to be checked in detail for their calculation methods and applicable scope.

[0076] If the to-be-verified three-level indicators have semantic cross or semantic contradiction, a multi-dimensional data verification mechanism such as comparative analysis and logic verification needs to be introduced to ensure the high accuracy and reliability of the final conclusion. In this process, the to-be-verified three-level indicators are compared with other indicators in multiple levels and multiple angles to find any possible inconsistencies. For example, in the personal credit history evaluation, the indicators about the overdue days under the credit card repayment record may need to be compared in detail with the indicators of the average monthly credit card usage limit to determine whether there is potential semantic overlap or conflict.

[0077] In summary, the social credit indicator library construction method based on big data according to the embodiments of the present application is illustrated, which extracts three-level indicator data sets from the social credit indicator library, and uses deep learning algorithm to perform semantic embedding coding on each three-level indicator to understand the semantic information of each three-level indicator. Then, after determining the to-be-verified three-level indicator, the to-be-verified three-level indicator is excluded from the data set, and a three-level indicator semantic environment is constructed based on the remaining three-level indicator data set. By performing dynamic semantic query response analysis on the to-be-verified three-level indicator in the constructed three-level indicator semantic environment, it is identified whether the to-be-verified three-level indicator has semantic cross or semantic contradiction, thereby realizing the deep verification of the three-level indicators. In this way, efficient and accurate verification of the three-level indicators can be realized, which helps to construct a more scientific and accurate social credit indicator library and improve the fairness and reliability of social credit evaluation.

[0078] Further, a social credit indicator library construction system based on big data is also provided.

[0079] Figure 6 The block diagram of the social credit indicator library construction system based on big data according to the embodiments of the present application is shown in FIG. 1. As shown in FIG. 1, the system includes a three-level indicator data set extraction module 101, a three-level indicator semantic embedding coding module 102, a three-level indicator semantic environment construction module 103, and a three-level indicator semantic cross or semantic contradiction identification module 104. Figure 6As shown, the big data-based social credit index library construction system 100 according to the embodiment of the present application comprises: a three-level index extraction module 110 configured to extract three-level indexes from a social credit index library to obtain a data subset of three-level indexes; a semantic embedding coding module 120 configured to perform semantic embedding coding on each three-level index in the data subset of three-level indexes to obtain a data subset of three-level index semantic embedding coding vectors; a to-be-verified index extraction module 130 configured to extract a three-level index semantic embedding coding vector of a three-level index to be verified from the data subset of three-level index semantic embedding coding vectors as a to-be-verified three-level index semantic embedding coding vector; and a semantic query depth verification module 140 configured to perform semantic query depth verification on the to-be-verified three-level index semantic embedding coding vector based on the data subset of three-level index semantic embedding coding vectors to obtain an index verification result, wherein the semantic query depth verification on the to-be-verified three-level index semantic embedding coding vector comprises: performing semantic query dynamic response analysis on the to-be-verified three-level index semantic embedding coding vector after deleting the to-be-verified three-level index semantic embedding coding vector from the data subset of three-level index semantic embedding coding vectors to obtain the index verification result.

[0080] Here, those skilled in the art can understand that the specific operations of each module in the above big data-based social credit index library construction system have been described in detail above with reference to the description of the big data-based social credit index library construction method of the embodiment of the present application, and therefore, the repeated description thereof will be omitted. Figures 1 to 5

[0081] The above describes the basic principles of the present application in combination with specific embodiments, but it should be pointed out that the advantages, advantages, effects, etc. mentioned in the present application are only examples and are not limited, and these advantages, advantages, effects, etc. cannot be considered as the must-have of each embodiment of the present application. In addition, the specific details of the above embodiments are only for the purpose of example and for the purpose of understanding, and are not limited to the must-have of the above specific details to realize the present application.

[0082] ​In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in a certain embodiment can be referred to the relevant description of other embodiments. In the several embodiments provided by the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are merely schematic, for example, the unit division is only a logical function division, and other division manners can be used in actual implementation. The units described as separated components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e. they can be located in one place or distributed on a plurality of network units. In actual implementation, some or all of the units can be selected according to the actual needs to achieve the purposes of the embodiments.

[0083] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and all changes falling within the meaning and range of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be considered as limiting the claims to which they relate.

[0084] Furthermore, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in a system claim can also be implemented by one unit by means of software or hardware.

[0085] Finally, it should be noted that the above description is given for the purpose of illustration and description. In addition, the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for constructing a social credit index database based on big data, characterized in that: include: Extracting the third-level indicators from the social credit indicator database to obtain a data subset of the third-level indicators; Performing semantic embedding coding on each of the three-level indicators in the data subset of the three-level indicators to obtain a data subset of the three-level indicator semantic embedding coding vector; Extracting the third-level indicator semantic embedding coding vector of the third-level indicator to be verified from the data subset of the third-level indicator semantic embedding coding vector as the third-level indicator semantic embedding coding vector to be verified; Based on the data subset of the three-level indicator semantic embedding coding vector, performing a semantic query deep verification on the three-level indicator semantic embedding coding vector to be verified to obtain an indicator verification result; Based on the data subset of the three-level indicator semantic embedding coding vector, a semantic query deep verification is performed on the three-level indicator semantic embedding coding vector to be verified to obtain an indicator verification result, including: Deleting the to-be-verified third-level indicator semantic embedding coding vector from the data subset of the third-level indicator semantic embedding coding vector to obtain a data set of the third-level indicator library semantic embedding coding vector; Performing a three-level indicator semantic query verification encoding based on adaptive weight anchoring on the data set of the semantic embedding coding vector of the third-level indicator to be verified and the semantic embedding coding vector of the third-level indicator library to obtain a third-level indicator semantic query deep verification coding vector to be verified; Based on the semantic query of the third-level indicator to be verified, the depth verification coding vector is determined, and the indicator verification result is used to indicate whether there is semantic overlap or semantic contradiction in the third-level indicator to be verified; Performing a three-level indicator semantic query verification encoding based on adaptive weight anchoring on the data set of the semantic embedding coding vector of the third-level indicator to be verified and the semantic embedding coding vector of the third-level indicator library to obtain a third-level indicator semantic query deep verification coding vector to be verified, including: Performing deep latent feature extraction on each of the three-level indicator library semantic embedding coding vectors in the data set of the three-level indicator semantic embedding coding vector to be verified and the three-level indicator library semantic embedding coding vector to obtain a data set of the three-level indicator semantic query feature deep latent coding vector to be verified and the three-level indicator library semantic feature deep latent coding vector; Performing semantic response decision anchor coding on each of the three-level indicator library semantic feature depth implicit coding vectors in the data set of the three-level indicator semantic query feature depth implicit coding vector to be verified and the three-level indicator library semantic feature depth implicit coding vector to obtain a data set of the three-level indicator to be verified-indicator library semantic response anchor coding matrix; Dynamically aggregate and encode the data set of the to-be-verified third-level indicator-indicator library semantic response anchor coding matrix to obtain the to-be-verified third-level indicator semantic query dynamic response coding vector.

2. The method for constructing a social credit index database based on big data according to claim 1 is characterized in that: Performing semantic embedding coding on each of the three-level indicators in the data subset of the three-level indicators to obtain a data subset of the three-level indicator semantic embedding coding vector, including: Each of the three-level indicators in the data subset of the three-level indicators is input into a semantic embedding encoder based on the Bert model to obtain a data subset of the three-level indicator semantic embedding coding vector.

3. The method for constructing a social credit index database based on big data according to claim 2 is characterized in that: The method further comprises performing deep latent feature extraction on each of the three-level indicator library semantic embedding coding vectors in the data set of the three-level indicator semantic embedding coding vector to be verified and the three-level indicator library semantic embedding coding vector to obtain a data set of the three-level indicator semantic query feature deep latent coding vector to be verified and the three-level indicator library semantic feature deep latent coding vector, including: A deep implicit feature extraction module based on a fully connected coding network is used to process each three-level indicator library semantic embedding coding vector in the data set of the three-level indicator semantic embedding coding vector to be verified and the three-level indicator library semantic embedding coding vector to obtain the data set of the three-level indicator semantic query feature deep implicit coding vector to be verified and the three-level indicator library semantic feature deep implicit coding vector.

4. The method for constructing a social credit index database based on big data according to claim 3 is characterized in that: Dynamically aggregate encoding is performed on the data set of the to-be-verified third-level indicator-indicator library semantic response anchor encoding matrix to obtain the to-be-verified third-level indicator semantic query dynamic response encoding vector, including: Based on the feature distribution of each of the three-level indicator-index library semantic response anchor coding matrices to be verified in the data set of the three-level indicator-index library semantic response anchor coding matrix to be verified, calculating the decision anchor adaptive splicing factor of each of the three-level indicator-index library semantic response anchor coding matrices to be verified to obtain a data set of the three-level indicator-index library decision anchor adaptive splicing factor to be verified, wherein the three-level indicator-index library decision anchor adaptive splicing factor to be verified is related to the maximum eigenvalue, characteristic variance, characteristic mean and number of eigenvalues ​​of the three-level indicator-index library semantic response anchor coding matrix to be verified; Performing a weighting process based on the Softmax function on the dataset of the three-level indicator to be verified-the indicator library decision anchor adaptive splicing factor to obtain a dataset of the three-level indicator to be verified-the indicator library decision anchor adaptive splicing weight factor; Based on the data set of the adaptive splicing weight factor of the decision anchor of the three-level indicator to be verified-the indicator library, the data set of the semantic response anchor coding matrix of the three-level indicator to be verified-the indicator library is weighted fused and feature-reshaped to obtain the dynamic response coding vector of the semantic query of the three-level indicator to be verified.

5. The method for constructing a social credit index database based on big data according to claim 4 is characterized in that: Based on the feature distribution of each of the three-level indicators to be verified - indicator library semantic response anchor coding matrices in the data set of the three-level indicators to be verified - indicator library semantic response anchor coding matrices, calculating the decision anchor adaptive splicing factor of each of the three-level indicators to be verified - indicator library semantic response anchor coding matrices to obtain a data set of the three-level indicators to be verified - indicator library decision anchor adaptive splicing factors, including: The sum of the characteristic variance and the drift coefficient of the semantic response anchor coding matrix of the three-level indicator-indicator library to be verified is used as the numerator, and the square of the difference between the maximum eigenvalue and the characteristic mean of the semantic response anchor coding matrix of the three-level indicator-indicator library to be verified is calculated and multiplied by the number of its eigenvalues, and then the drift coefficient and twice the characteristic variance are added as the denominator to obtain the adaptive splicing factor of the decision anchor of the three-level indicator-indicator library to be verified, wherein the drift coefficient is used to smooth the characteristic distribution fluctuation of the semantic response anchor coding matrix of the three-level indicator-indicator library to be verified.

6. The method for constructing a social credit index database based on big data according to claim 5 is characterized in that: Determining the indicator verification result based on the semantic query of the third-level indicator to be verified, including: The semantic query depth verification coding vector of the third-level indicator to be verified is input into the classifier-based indicator verifier to obtain the indicator verification result.

7. The method for constructing a social credit index database based on big data according to claim 6, characterized in that: Inputting the to-be-verified third-level indicator semantic query depth verification encoding vector into the classifier-based indicator verifier to obtain the indicator verification result, including: Using the fully connected layer of the indicator checker to perform fully connected encoding on the three-level indicator semantic query deep verification code vector to be verified to obtain the three-level indicator semantic query deep verification fully connected code vector to be verified; Inputting the to-be-verified three-level indicator semantic query deep verification fully connected code vector into the Softmax classification function of the indicator verifier to obtain probability values ​​of the to-be-verified three-level indicator semantic query deep verification code vector belonging to each classification label, wherein the classification label includes whether the to-be-verified three-level indicator semantic query deep verification code vector has semantic intersection or semantic contradiction and whether it does not have semantic intersection or semantic contradiction; The classification label corresponding to the largest probability value among the probability values ​​is determined as the indicator verification result.

8. A system for constructing a social credit index database based on big data, used to execute the method according to any one of claims 1 to 7, characterized in that: include: A third-level indicator extraction module is used to extract third-level indicators from the social credit indicator database to obtain a data subset of the third-level indicators; A semantic embedding coding module, configured to perform semantic embedding coding on each of the three-level indicators in the data subset of the three-level indicators to obtain a data subset of the three-level indicator semantic embedding coding vector; The indicator extraction module to be verified is used to extract the third-level indicator semantic embedding coding vector of the third-level indicator to be verified from the data subset of the third-level indicator semantic embedding coding vector as the third-level indicator semantic embedding coding vector to be verified; The semantic query deep verification module is used to perform semantic query deep verification on the semantic embedding coding vector of the third-level indicator to be verified based on the data subset of the third-level indicator semantic embedding coding vector to obtain the indicator verification result.

Images