Credit enhancement evaluation system based on knowledge graph and deep learning

Abstract

The invention relates to the technical field of user data management, and particularly discloses a knowledge graph and deep learning-based credit enhancement evaluation system, which comprises a personal credit rating calculation module, a credit evaluation module and a credit evaluation module, and is characterized in that the personal credit rating calculation module calculates a current personal credit score value based on a knowledge graph and deep learning method; if the first credit enhancement evaluation execution module judges that the credit score value of the current individual is not lower than the score value threshold value corresponding to the target bank lending limit, the credit score value of the current individual is used as a credit enhancement evaluation result; if the second credit increase evaluation execution module judges that the credit score value of the current individual is lower than the score value threshold value corresponding to the target bank loan limit, the second credit increase evaluation execution module calculates a family credit score value based on the credit related index information of the family members of the current individual; obtaining a credit increase evaluation result of the current person based on the credit score value of the current person and the household credit score value; the method and the device are used for realizing accurate and comprehensive credit enhancement evaluation of personal credit conditions by utilizing the credit conditions of personal family members.

Description

Technical Field

[0001] The present invention relates to the technical field of user data management, and in particular to a credit enhancement evaluation system based on knowledge graph and deep learning. Background Art

[0002] In today's financial field, credit assessment is crucial for banks and other financial institutions to make lending decisions. With the continuous development and innovation of the financial market, traditional credit assessment methods have gradually revealed their limitations. In order to more accurately and comprehensively assess the credit status of individuals, a credit enhancement assessment system based on knowledge graphs and deep learning has emerged. Knowledge graphs can integrate a large amount of structured and unstructured data, build a complex relationship network, and clearly show the relationship between various information related to personal credit. Deep learning methods have powerful learning and prediction capabilities, and can mine hidden patterns and features from massive data. By combining the two, a person's credit status can be analyzed more deeply. In practical applications, an individual's credit status is affected by many factors, including not only their own financial status, credit history, etc., but also may be related to the credit status of family members.

[0003] However, the existing credit assessment system often only focuses on an individual's independent credit data, and does not adequately consider related factors such as family credit, making it difficult to comprehensively and accurately assess an individual's credit level.

[0004] Therefore, the present invention proposes a credit enhancement evaluation system based on knowledge graph and deep learning. Summary of the invention

[0005] The present invention provides a credit enhancement assessment system based on knowledge graph and deep learning, which is used to use the personal credit rating calculation module to calculate the credit score value using the knowledge graph and deep learning method, thereby improving the accuracy and scientificity of the score. The first credit enhancement assessment execution module can quickly determine that individuals with qualified credit scores do not need credit enhancement assessment, thereby improving the assessment efficiency. For individuals whose credit scores do not meet the standards, the second credit enhancement assessment execution module screens the credit indicators of family members and calculates the family credit score value, and then calculates the family credit enhancement score value according to the weight in combination with the personal credit score value, comprehensively considering the impact of family factors on personal credit and broadening the assessment dimension. On the whole, the system can accurately and comprehensively assess the credit status of individuals, provide a reliable basis for banks' lending decisions, and at the same time improve the efficiency and accuracy of assessments and reduce credit risks.

[0006] The present invention provides a credit enhancement evaluation system based on knowledge graph and deep learning, comprising:

[0007] Personal credit rating calculation module, used to calculate the current personal credit score based on knowledge graph and deep learning methods;

[0008] The first credit enhancement evaluation execution module is used to determine that the current individual does not need credit enhancement evaluation if it is determined that the current individual's credit score is not lower than the score threshold corresponding to the target bank's loan limit, and use the current individual's credit score as the current individual's credit enhancement evaluation result;

[0009] The second credit enhancement assessment execution module is used to determine that the current individual needs a credit enhancement assessment if it is determined that the current individual's credit score is lower than the score threshold corresponding to the target bank's loan amount, and to screen out the high-quality indicator values ​​of each secondary indicator under each primary indicator from the credit-related indicator information of the current individual's family members, and calculate the current individual's family credit score based on the high-quality indicator values ​​of all secondary indicators under all primary indicators, and calculate the current individual's credit score and family credit score by weight to obtain the current individual's family credit enhancement score as the current individual's credit enhancement assessment result.

[0010] Preferably, it also includes:

[0011] The first credit approval module is used to input the current individual's credit score value into the loan process processing end if the current individual's credit enhancement assessment result is the current individual's credit score value;

[0012] The second credit approval module is used to determine whether the current individual's family credit enhancement score is not lower than the score threshold corresponding to the target bank's loan amount if the current individual's credit enhancement assessment result is the current individual's family credit enhancement score. If so, the current individual's family credit enhancement score is input into the loan process processing end; otherwise, the current individual's family credit enhancement score is stored in the cloud.

[0013] Preferably, the personal credit rating calculation module includes:

[0014] A credit-related information acquisition submodule, used to collect the credit-related information of the current individual and the credit-related information of multiple reference individuals from multiple data sources;

[0015] A knowledge graph generation submodule, for generating a knowledge graph of the credit index items of the current individual and a knowledge graph of the credit index items of each reference individual based on the credit-related information of the current individual and the credit-related information of all reference individuals;

[0016] The credit scoring model building submodule is used to build a personal credit scoring model based on the step-by-step loan repayment requirements and the knowledge graph of the original credit index items of all reference individuals;

[0017] The personal credit scoring submodule is used to input the current individual's credit index item knowledge graph into the personal credit scoring model to obtain the current individual's credit score value.

[0018] Preferably, the knowledge graph generation submodule includes:

[0019] A credit-related information interpretation unit, used to identify all valid credit index items of all personal credit-related information and determine the inference logic between all valid credit index items;

[0020] A knowledge graph framework generation unit is used to treat each credit index item as an entity, and construct a knowledge graph framework for all credit index items based on the inference logic between all credit index items, so as to obtain the original credit index item knowledge graph framework of each individual;

[0021] The entity attribute value normalization unit is used to normalize the performance value of the same entity in the original credit index item knowledge graph framework of all individuals in the credit-related information of all individuals, and obtain the normalized value of each entity in the credit-related information of each individual;

[0022] A knowledge graph framework assignment unit, used to assign a value to the credit index item knowledge graph framework of each individual based on the normalized value of each entity in the credit-related information of each individual, so as to obtain the credit index item knowledge graph of each individual;

[0023] Among them, the knowledge graph of credit index items of all individuals includes the knowledge graph of credit index items of the current individual and the knowledge graph of credit index items of all reference individuals.

[0024] Preferably, the credit-related information interpretation unit includes:

[0025] The inference logic generation subunit is used to determine all characteristic attributes of each piece of information in all personal credit-related information, trace back the numerical inference logic of the characteristic attributes of all pieces of information in each personal credit-related information, and obtain the inference logic between all characteristic attributes;

[0026] A marginal feature attribute screening unit, used to treat the feature attributes that do not involve any inference logic among all the feature attributes of each individual's credit-related information as marginal feature attributes;

[0027] A contribution rate calculation unit is used to calculate the contribution rate of all marginal feature attributes in all personal credit related information;

[0028] The effective index item screening unit is used to treat all the marginal characteristic attributes whose contribution rate in all personal credit-related information is not less than the contribution rate threshold and all the characteristic attributes involved in the inference logic in the corresponding credit-related information as the effective credit index items in the corresponding credit-related information;

[0029] An upper-layer inference logic evolution unit, used to perform upper-layer inference logic evolution on each valid credit index item not involving inference logic, and obtain an upper-layer inference logic evolution result of each valid credit index item not involving inference logic;

[0030] The inference logic merging unit is used to merge the upper-level inference logic evolution results of all valid credit index items that do not involve the inference logic with the inference logic between all feature attributes to obtain the inference logic between all valid credit index items.

[0031] Preferably, the upper layer inference logic evolution unit includes:

[0032] The co-occurrence count subunit is used to obtain a large amount of credit-related information, treat each valid credit index item not involving inference logic as a marginal valid credit index item, and count the co-occurrence times of each marginal valid credit index item and each valid credit index item involving inference logic in all credit-related information;

[0033] A co-occurrence frequency calculation subunit, used to take the quotient of the sum of the number of characters of each marginal effective credit index item and each effective credit index item involved in the inference logic and the corresponding number of co-occurrences in all copies of credit-related information and the sum of the number of characters in all copies of credit-related information as the co-occurrence frequency of each marginal effective credit index item and each effective credit index item involved in the inference logic in all copies of credit-related information;

[0034] A credit index item screening subunit is used to use all valid credit index items involved in the inference logic with a co-occurrence frequency of each marginal valid credit index item in all credit-related data that is not less than a preset co-occurrence frequency as the basis credit index items of each marginal valid credit index item;

[0035] The upper-level inference logic evolution subunit is used to mine the inference logic between each marginal effective credit indicator item and all corresponding based credit indicator items as the upper-level inference logic evolution result of each marginal effective credit indicator item based on all co-occurring sentences between each marginal effective credit indicator item and all corresponding based credit indicator items in all credit-related materials.

[0036] Preferably, the credit scoring model building submodule includes:

[0037] The excellent sample screening submodule is used to screen out excellent individual sample groups for each loan amount value in the ladder loan amount from all reference individuals based on the ladder loan repayment requirements;

[0038] The loan quota and credit score value alignment submodule is used to align each loan quota value with the enterprise's loan quota credit score benchmark to obtain the standard credit score value corresponding to each loan quota value;

[0039] The credit scoring model building submodule is used to conduct deep learning and model building on the standard credit scoring value corresponding to each loan limit value and the original credit index item knowledge graph of all reference individuals in the corresponding excellent individual sample group, so as to construct a personal credit scoring model.

[0040] Preferably, the second credit enhancement assessment execution module includes:

[0041] The family member credit index information acquisition submodule is used to determine that the current individual needs credit enhancement assessment and obtain the credit-related index information of the current individual's family members if it is determined that the current individual's credit score is lower than the score threshold corresponding to the target bank's loan amount;

[0042] The high-quality index value screening submodule is used to determine the score value of each secondary index under each primary index for each family member based on the credit-related index information of the current individual's family members and the preset family credit index evaluation standard, and screen out the maximum score value from the score values ​​of each secondary index under each primary index of all family members as the high-quality index value of the current individual's family members under the corresponding secondary index under the corresponding primary index;

[0043] A family credit score calculation submodule is used to calculate the family credit score of the current individual based on the high-quality index values ​​of each secondary index under each primary index of the family members of the current individual;

[0044] The family credit enhancement score calculation submodule is used to calculate the current individual's credit score and the family credit score according to the weight, and obtain the current individual's family credit enhancement score as the current individual's credit enhancement assessment result.

[0045] Preferably, the method for the family credit score calculation submodule to calculate the family credit score of the current individual based on the high-quality index value of each secondary index under each primary index of the family members of the current individual includes:

[0046]

[0047] In the formula, Z k is the credit score value of the kth primary indicator, q k is the total number of all secondary indicators under the kth primary indicator, Z k,x is the quality index value of the current individual’s family member under the xth secondary index under the kth primary index, W k,x is the preset weight of the xth secondary indicator under the kth primary indicator, M is the current individual's family credit score, p is the total number of all primary indicators, W k is the preset weight of the kth first-level indicator.

[0048] Preferably, the family credit enhancement score value calculation submodule includes:

[0049] The family credit enhancement score calculation unit is used to calculate the current individual credit score and the family credit score according to the weight to obtain the current individual family credit enhancement score:

[0050] F=PW p +MW m

[0051] In the formula, F is the current individual's family credit enhancement score, P is the current individual's credit score, and W p is the preset weight of the current individual's credit score, M is the current individual's family credit score, and W m The preset weight for the current individual's household credit score value;

[0052] The credit enhancement evaluation result output unit is used to use the current individual's family credit enhancement score value as the current individual's credit enhancement evaluation result.

[0053] The beneficial effects of the present invention compared to the prior art are as follows: the credit score value is calculated using the knowledge graph and deep learning method using the personal credit rating calculation module, thereby improving the accuracy and scientificity of the score. The first credit enhancement assessment execution module can quickly determine that individuals with qualified credit scores do not need credit enhancement assessment, thereby improving the assessment efficiency. For individuals whose credit scores do not meet the standards, the second credit enhancement assessment execution module screens the credit indicators of family members and calculates the family credit score value, and then calculates the family credit enhancement score value by weight in combination with the personal credit score value, comprehensively considering the impact of family factors on personal credit and broadening the assessment dimension. On the whole, the system can accurately and comprehensively assess the credit status of individuals, provide a reliable basis for banks' lending decisions, and at the same time improve the efficiency and accuracy of assessments and reduce credit risks.

[0054] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in this application document.

[0055] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0057] Figure 1Schematic diagram of a credit enhancement assessment system based on knowledge graph and deep learning in an embodiment of the present invention;

[0058] Figure 2 This is a flow chart of a family credit enhancement algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0060] Embodiment 1:

[0061] The present invention provides a credit enhancement evaluation system based on knowledge graph and deep learning. Figure 1 and 2 ,include:

[0062] Personal credit rating calculation module, used to calculate the current personal credit score based on knowledge graph and deep learning methods;

[0063] The first credit enhancement evaluation execution module is used to determine that the current individual does not need credit enhancement evaluation if it is determined that the current individual's credit score is not lower than the score threshold corresponding to the target bank's loan limit, and use the current individual's credit score as the current individual's credit enhancement evaluation result;

[0064] The second credit enhancement assessment execution module is used to determine that the current individual needs a credit enhancement assessment if it is determined that the current individual's credit score is lower than the score threshold corresponding to the target bank's loan amount, and to screen out the high-quality indicator values ​​of each secondary indicator under each primary indicator from the credit-related indicator information of the current individual's family members, and calculate the current individual's family credit score based on the high-quality indicator values ​​of all secondary indicators under all primary indicators, and calculate the current individual's credit score and family credit score by weight to obtain the current individual's family credit enhancement score as the current individual's credit enhancement assessment result.

[0065] In this embodiment, the current individual credit score value is a value obtained by evaluating the current individual credit status through a specific calculation method and model.

[0066] In this embodiment, the target bank loan amount is the bank loan amount that the individual currently wants to apply for.

[0067] In this embodiment, the score value threshold corresponding to the target bank's loan amount is the minimum value that the personal credit score needs to reach in order to obtain the target bank's loan amount.

[0068] In this embodiment, the credit-related index information of family members is relevant data that can reflect the credit level of family members in terms of basic situation, financial situation, credit situation, social relationship, etc.

[0069] In this embodiment, the first-level indicators are relatively high-level and general classification indicators in the credit evaluation system, such as basic information, financial status, etc.

[0070] In this embodiment, all secondary indicators of the primary indicator are further subdivisions and specificities of the primary indicator, such as secondary indicators such as health status and marital status under the primary indicator of basic situation.

[0071] In this embodiment, the high-quality index value is the highest value among the scores of each secondary index under each primary index among the credit-related indexes of the family members.

[0072] In this embodiment, the current individual family credit score is a score reflecting the overall credit status of the family calculated based on the high-quality indicator values ​​of the secondary indicators of each primary indicator of the family members according to certain weights.

[0073] In this embodiment, the current individual's family credit enhancement score is a score obtained by calculating the current individual's credit score and the family's credit score according to their respective preset weights.

[0074] The beneficial effects of the above technology are: the personal credit rating calculation module uses knowledge graphs and deep learning methods to calculate credit scores, which improves the accuracy and scientificity of the scores. The first credit enhancement assessment execution module can quickly determine that individuals with qualified credit scores do not need credit enhancement assessments, thereby improving assessment efficiency. For individuals whose credit scores do not meet the standards, the second credit enhancement assessment execution module screens the credit indicators of family members and calculates the family credit score, and then calculates the family credit enhancement score by weight based on the personal credit score, comprehensively considering the impact of family factors on personal credit and broadening the assessment dimension. Overall, the system can accurately and comprehensively assess an individual's credit status, provide a reliable basis for banks' lending decisions, and at the same time improve assessment efficiency and accuracy, and reduce credit risk.

[0075] Embodiment 2:

[0076] On the basis of Example 1, it also includes:

[0077] The first credit approval module is used to input the current individual's credit score value into the loan process processing end if the current individual's credit enhancement assessment result is the current individual's credit score value;

[0078] The second credit approval module is used to determine whether the current individual's family credit enhancement score is not lower than the score threshold corresponding to the target bank's loan amount if the current individual's credit enhancement assessment result is the current individual's family credit enhancement score. If so, the current individual's family credit enhancement score is input into the loan process processing end; otherwise, the current individual's family credit enhancement score is stored in the cloud.

[0079] In this embodiment, the loan process processing end refers to the system, department or link responsible for processing and executing bank loan related operations and processes. It covers a series of tasks such as receiving and reviewing personal credit assessment results, and finally deciding whether to issue a loan and specific loan operations.

[0080] The beneficial effects of the above technology are as follows: When the credit enhancement assessment result is a personal credit score value, the first credit approval module will put it into the loan process processing end, simplifying the approval process and improving processing efficiency. The second credit approval module further judges the family credit enhancement score value. If it meets the standard, it will be put into the loan process processing end. If it does not meet the standard, it will be stored in the cloud, realizing the classification processing of different assessment results and ensuring the accuracy and rationality of the approval. Overall, the setting of these two modules improves the credit approval process, improves the efficiency and accuracy of the approval, can more effectively manage and process personal credit assessment results, and provide more orderly and reliable support for the bank's lending decisions.

[0081] Embodiment 3:

[0082] Based on Example 1, the personal credit rating calculation module includes:

[0083] A credit-related information acquisition submodule, used to collect the credit-related information of the current individual and the credit-related information of multiple reference individuals from multiple data sources;

[0084] A knowledge graph generation submodule, for generating a knowledge graph of the credit index items of the current individual and a knowledge graph of the credit index items of each reference individual based on the credit-related information of the current individual and the credit-related information of all reference individuals;

[0085] The credit scoring model building submodule is used to build a personal credit scoring model based on the step-by-step loan repayment requirements and the knowledge graph of the original credit index items of all reference individuals;

[0086] The personal credit scoring submodule is used to input the current individual's credit index item knowledge graph into the personal credit scoring model to obtain the current individual's credit score value.

[0087] In this embodiment, collecting the credit-related information of the current individual and the credit-related information of multiple reference individuals from multiple data sources means obtaining the credit-related data and information of the individual currently applying for credit assessment and other individuals used for comparison and reference from various different channels.

[0088] In this embodiment, the reference individuals are representative individual samples used to construct and calibrate the credit assessment model, and their credit-related information can be used as a standard for comparison and reference.

[0089] In this embodiment, the credit-related information includes various data and situations of an individual, such as his / her basic situation, financial status, credit history, social relations, etc., which can reflect his / her credit level and debt repayment ability.

[0090] In this embodiment, the credit index item knowledge graph is a structure that displays and associates various indicators and elements related to personal credit in the form of a graph, which is used to present and analyze the credit status more clearly.

[0091] In this embodiment, the step-by-step loan repayment requirement may refer to phased or hierarchical repayment regulations and requirements set according to different loan amounts, terms, interest rates and other conditions.

[0092] In this embodiment, the personal credit scoring model is a mathematical model constructed by using data and algorithms to quantitatively evaluate and score personal credit status.

[0093] The beneficial effects of the above technologies are as follows: the credit-related information acquisition submodule collects information from multiple data sources, ensuring the comprehensiveness and diversity of the data. The knowledge graph generation submodule generates a knowledge graph of personal credit index items, presenting credit information in a more intuitive and clear way. The credit scoring model construction submodule builds a model based on the step-by-step loan repayment requirements and the reference individual's knowledge graph, which improves the model's pertinence and adaptability. The personal credit scoring submodule inputs the current individual's knowledge graph into the model to obtain a credit score value, ensuring the accuracy and objectivity of the score. Overall, the personal credit rating calculation module can comprehensively and accurately calculate the current individual's credit score value through the collaborative work of these submodules, providing a reliable foundation for subsequent credit enhancement assessment and credit approval, and helping to improve the performance and effectiveness of the entire credit enhancement assessment system.

[0094] Embodiment 4:

[0095] Based on Example 3, the knowledge graph generation submodule includes:

[0096] A credit-related information interpretation unit, used to identify all valid credit index items of all personal credit-related information and determine the inference logic between all valid credit index items;

[0097] A knowledge graph framework generation unit is used to treat each credit index item as an entity, and construct a knowledge graph framework for all credit index items based on the inference logic between all credit index items, so as to obtain the original credit index item knowledge graph framework of each individual;

[0098] The entity attribute value normalization unit is used to normalize the performance value of the same entity in the original credit index item knowledge graph framework of all individuals in the credit-related information of all individuals, and obtain the normalized value of each entity in the credit-related information of each individual;

[0099] A knowledge graph framework assignment unit, used to assign a value to the credit index item knowledge graph framework of each individual based on the normalized value of each entity in the credit-related information of each individual, so as to obtain the credit index item knowledge graph of each individual;

[0100] Among them, the knowledge graph of credit index items of all individuals includes the knowledge graph of credit index items of the current individual and the knowledge graph of credit index items of all reference individuals.

[0101] In this embodiment, the effective credit index items are index items in the personal credit related information that have been screened and calculated and have a practical role and significant impact on the credit assessment.

[0102] In this embodiment, determining the inference logic between all valid credit index items is to find out the rules of inference and association between these valuable credit index items. The specific implementation process includes:

[0103] For each valid credit indicator item, analyze its appearance and value changes in different credit-related information;

[0104] Through statistical analysis and data mining techniques, explore the correlation and dependency between these indicators;

[0105] For example, observe whether the change of a certain indicator item will lead to corresponding changes in other indicators, or whether certain indicators always appear at the same time or have a specific sequential relationship;

[0106] Machine learning algorithms can also be used to train and learn these data to automatically discover potential inference patterns;

[0107] At the same time, the preliminary inference logic discovered is verified and revised by combining domain knowledge and expert experience;

[0108] Finally, by integrating all analysis and verification results, a clear, accurate and practical inference logic between all valid credit indicator items is determined.

[0109] In this embodiment, the original credit index item knowledge graph framework is an initially constructed, unprocessed and unperfected graph structure used to represent the relationship between credit index items.

[0110] In this embodiment, the performance values ​​of the same entity in the original credit index item knowledge graph framework of all individuals in the credit-related information of all individuals are normalized to obtain the normalized value of each entity in the credit-related information of each individual, that is, the specific numerical values ​​corresponding to the same credit index items of different individuals in the original knowledge graph framework are standardized by a certain method to obtain a unified numerical value within a specific range for comparison and analysis.

[0111] The beneficial effects of the above technologies are as follows: the credit-related information interpretation unit identifies valid credit indicator items and inference logic, lays the foundation for building a knowledge graph, and improves the accuracy and effectiveness of information. The knowledge graph framework generation unit constructs the original knowledge graph framework to make the relationship between credit indicator items clearer and more organized. The entity attribute value normalization unit normalizes the entity performance value to ensure the comparability and consistency of the data. The knowledge graph framework assignment unit assigns values ​​to the knowledge graph framework to obtain a complete credit indicator item knowledge graph, which more accurately reflects the individual's credit status. On the whole, the knowledge graph generation submodule can generate high-quality, accurate and comparable credit indicator item knowledge graphs through the collaborative work of these units, providing a solid data foundation for subsequent credit scoring and evaluation, and helping to improve the accuracy and reliability of the credit enhancement evaluation system.

[0112] Embodiment 5:

[0113] Based on the fourth embodiment, the credit-related information interpretation unit includes:

[0114] The inference logic generation subunit is used to determine all characteristic attributes of each piece of information in all personal credit-related information, trace back the numerical inference logic of the characteristic attributes of all pieces of information in each personal credit-related information, and obtain the inference logic between all characteristic attributes;

[0115] A marginal feature attribute screening unit, used to treat the feature attributes that do not involve any inference logic among all the feature attributes of each individual's credit-related information as marginal feature attributes;

[0116] A contribution rate calculation unit is used to calculate the contribution rate of all marginal feature attributes in all personal credit related information;

[0117] The effective index item screening unit is used to treat all the marginal characteristic attributes whose contribution rate in all personal credit-related information is not less than the contribution rate threshold and all the characteristic attributes involved in the inference logic in the corresponding credit-related information as the effective credit index items in the corresponding credit-related information;

[0118] An upper-layer inference logic evolution unit, used to perform upper-layer inference logic evolution on each valid credit index item not involving inference logic, and obtain an upper-layer inference logic evolution result of each valid credit index item not involving inference logic;

[0119] The inference logic merging unit is used to merge the upper-level inference logic evolution results of all valid credit index items that do not involve the inference logic with the inference logic between all feature attributes to obtain the inference logic between all valid credit index items.

[0120] In this embodiment, each piece of information in the credit-related information refers to a separate and specific piece of content in a large amount of information related to personal credit.

[0121] In this embodiment, all characteristic attributes of each piece of information are various properties contained in each piece of credit-related information that can describe and reflect its characteristics.

[0122] In this embodiment, the numerical calculation logic of the characteristic attributes of all pieces of information related to each individual's credit is traced back to obtain the calculation logic between all characteristic attributes, that is, to deeply explore how the numerical values ​​of the characteristic attributes in all information related to each person's credit are derived and associated, so as to derive the internal laws between these characteristic attributes.

[0123] In this embodiment, the contribution rate of all marginal feature attributes in the credit-related information of all individuals is calculated: the role or contribution of those feature attributes in a secondary or non-core position in the credit information of all individuals is calculated. The calculation process includes:

[0124] Count the number of times edge feature attributes appear in all credit-related information;

[0125] Calculate the total amount of credit-related information for all individuals;

[0126] Divide the number of occurrences of the edge feature attribute by the total amount of credit-related information to obtain the frequency of occurrence of the edge feature attribute;

[0127] The degree of influence of edge feature attributes on the credit assessment results can also be further considered. For example, if the existence or value change of a certain edge feature attribute has a greater impact on the final credit score, a higher weight can be given to calculate its contribution rate;

[0128] Taking into account factors such as frequency and impact, the contribution rate of each edge feature attribute to all personal credit-related information is finally obtained.

[0129] In this embodiment, the contribution rate threshold is a pre-set contribution rate numerical standard used to determine whether the edge feature attribute is sufficiently important.

[0130] In this embodiment, the upper-level inference logic evolution results of all valid credit index items that do not involve inference logic are merged with the inference logic between all characteristic attributes to obtain the inference logic between all valid credit index items, that is, the logic obtained by further deducing those valid credit index items that originally did not have direct inference logic is combined with the original inference logic between all characteristic attributes, thereby obtaining a complete and comprehensive inference logic between all valid credit index items.

[0131] The beneficial effects of the above technology are as follows: the inference logic generation subunit traces the numerical inference logic of the feature attributes, which helps to fully understand the internal relationship of the credit-related information. The marginal feature attribute screening unit finds out the feature attributes that do not involve the inference logic, which provides a basis for subsequent analysis. The contribution rate calculation unit calculates the contribution rate of the marginal feature attributes and can measure their importance. The effective indicator item screening unit selects the marginal feature attributes with higher contribution rates and the feature attributes involving the inference logic as effective credit indicator items, thereby improving the effectiveness of the indicators. The upper-level inference logic evolution unit performs upper-level inference logic evolution for the effective credit indicator items that do not involve the inference logic, thereby improving the logic system. The inference logic merging unit merges different inference logics to form a complete inference logic between effective credit indicator items, making the credit assessment more comprehensive and accurate. On the whole, the credit-related information interpretation unit can accurately determine the effective credit indicator items and their inference logic in the credit-related information through the collaborative work of these subunits, providing key support for generating high-quality knowledge graphs, thereby improving the accuracy and reliability of the credit enhancement assessment system.

[0132] Embodiment 6:

[0133] On the basis of Embodiment 5, the upper layer inference logic evolution unit includes:

[0134] The co-occurrence count subunit is used to obtain a large amount of credit-related information, treat each valid credit index item not involving inference logic as a marginal valid credit index item, and count the co-occurrence times of each marginal valid credit index item and each valid credit index item involving inference logic in all credit-related information;

[0135] A co-occurrence frequency calculation subunit, used to take the quotient of the sum of the number of characters of each marginal effective credit index item and each effective credit index item involved in the inference logic and the corresponding number of co-occurrences in all copies of credit-related information and the sum of the number of characters in all copies of credit-related information as the co-occurrence frequency of each marginal effective credit index item and each effective credit index item involved in the inference logic in all copies of credit-related information;

[0136] A credit index item screening subunit is used to use all valid credit index items involved in the inference logic with a co-occurrence frequency of each marginal valid credit index item in all credit-related data that is not less than a preset co-occurrence frequency as the basis credit index items of each marginal valid credit index item;

[0137] The upper-level inference logic evolution subunit is used to mine the inference logic between each marginal effective credit indicator item and all corresponding based credit indicator items as the upper-level inference logic evolution result of each marginal effective credit indicator item based on all co-occurring sentences between each marginal effective credit indicator item and all corresponding based credit indicator items in all credit-related materials.

[0138] In this embodiment, the number of co-occurrences of each marginal effective credit index item and each effective credit index item involving inference logic in all credit-related materials refers to the number of times that a single marginal effective credit index item and those effective credit index items involving inference logic co-occur in a large amount of credit-related materials. This can be done in the following steps:

[0139] For each piece of information, identify the marginal effective credit index items and the effective credit index items involving the inference logic one by one;

[0140] During the identification process, whenever a marginal valid credit index item and a valid credit index item involving inference logic are found to appear in the same document at the same time, the corresponding co-occurrence count is increased by 1;

[0141] After completing the traversal and statistics of all the data, the number of co-occurrences of each edge valid credit index item and each valid credit index item involved in the inference logic in all the credit-related data is obtained.

[0142] This process may require the use of natural language processing technology and related data processing tools to improve efficiency and accuracy.

[0143] In this embodiment, the preset co-occurrence frequency is a frequency standard set in advance for determining the degree of association between the marginal effective credit index item and other basis credit index items.

[0144] In this embodiment, based on all co-occurring sentences between each marginal effective credit indicator item and all corresponding based credit indicator items in all credit-related materials, the inference logic between each marginal effective credit indicator item and all corresponding based credit indicator items is mined, that is, by analyzing all related sentences containing each marginal effective credit indicator item and all its corresponding based credit indicator items appearing at the same time, the deduction and association rules between them are found out.

[0145] The following steps can be taken:

[0146] Collect and organize all statements containing marginal effective credit index items and corresponding basis credit index items;

[0147] Perform text preprocessing on these co-occurring sentences, including noise cleaning, word segmentation, part-of-speech tagging, etc., for subsequent analysis;

[0148] Use natural language processing techniques and machine learning algorithms, such as association rule mining and dependency syntax analysis, to explore the relationships and structures between words in a sentence;

[0149] By analyzing the context information of the marginal effective credit index items and the credit index items in the co-occurring sentences, we can observe their positions, the words before and after them, and the overall semantic relationship.

[0150] Based on these analysis results, possible patterns and rules can be summarized and generalized to infer the logical relationship between the two. For example, if in multiple co-occurring sentences, it is found that the marginal effective credit indicator item "low income" always appears at the same time as the credit indicator item "high debt risk", and their contexts all mention "repayment difficulties", then it can be inferred that "low income" may lead to "high debt risk" and there is a certain causal relationship;

[0151] Continuously verify and optimize the mined inference logic and ensure its accuracy and effectiveness by comparing it with actual credit data and business experience.

[0152] The beneficial effects of the above technologies are as follows: the co-occurrence count subunit provides basic data for subsequent analysis by counting the co-occurrence counts of valid credit indicator items. The co-occurrence frequency calculation subunit calculates the co-occurrence frequency, which more accurately reflects the degree of correlation between indicator items. The credit indicator item screening subunit screens out the credit indicator items whose co-occurrence frequencies meet the requirements, thereby improving the accuracy of screening. The upper-level inference logic evolution subunit mines the inference logic as the result of the upper-level inference logic evolution, and improves the logical system of the credit indicator items. On the whole, the upper-level inference logic evolution unit, through the collaborative work of these subunits, can effectively evolve the upper-level inference logic for valid credit indicator items that do not involve inference logic, making the interpretation of credit-related information more comprehensive and in-depth, providing strong support for generating accurate and comprehensive credit assessment results, and improving the scientificity and reliability of the credit enhancement assessment system.

[0153] Embodiment 7:

[0154] Based on Example 3, the credit scoring model building submodule includes:

[0155] The excellent sample screening submodule is used to screen out excellent individual sample groups for each loan amount value in the ladder loan amount from all reference individuals based on the ladder loan repayment requirements;

[0156] The loan quota and credit score value alignment submodule is used to align each loan quota value with the enterprise's loan quota credit score benchmark to obtain the standard credit score value corresponding to each loan quota value;

[0157] The credit scoring model building submodule is used to conduct deep learning and model building on the standard credit scoring value corresponding to each loan limit value and the original credit index item knowledge graph of all reference individuals in the corresponding excellent individual sample group, so as to construct a personal credit scoring model.

[0158] In this embodiment, the tiered loan amount refers to the amount of loan funds divided into different levels or stages.

[0159] In this embodiment, the loan limit value is a specific amount of loan funds.

[0160] In this embodiment, the sample group of excellent individuals with loan limit values ​​is a group of individuals with good credit performance under a specific loan limit value.

[0161] In this embodiment, based on the step-by-step loan repayment requirements, a sample group of excellent individuals for each loan amount value in the step-by-step loan amount is screened out from all reference individuals. According to the phased loan repayment regulations and requirements, a group of individuals with good credit corresponding to each different loan amount value is selected from all reference individuals.

[0162] In this embodiment, each loan limit value is aligned with the enterprise's loan limit credit score benchmark, and the standard credit score value corresponding to each loan limit value is obtained as follows:

[0163] Compare each specific loan limit value with the credit scoring standard set by the enterprise for the loan limit, so as to obtain the standard credit score corresponding to this loan limit value.

[0164] In this embodiment, deep learning and model building are performed on the standard credit score value corresponding to each loan limit value and the original credit index item knowledge graph of all reference individuals in the corresponding excellent individual sample group to construct a personal credit score model:

[0165] Using the standard credit score value for each loan limit value and the original credit index knowledge graph of the reference individuals in the corresponding excellent individual sample group, a personal credit score model is established through deep learning methods. In this model construction process, the model input is the original credit index item knowledge graph of each reference individual in the excellent individual sample group corresponding to each loan limit value, and the output is the standard credit score value corresponding to each loan limit value.

[0166] The beneficial effects of the above technologies are as follows: the excellent sample screening submodule selects excellent individual sample groups based on the step-by-step loan repayment requirements, ensuring the quality and representativeness of the samples. The loan limit and credit score value alignment submodule aligns the loan limit value with the standard credit score value, providing clear goals and standards for model building. The credit score model building submodule constructs a more accurate and reliable personal credit score model through deep learning and model building of the knowledge graph of the standard credit score value and the excellent individual sample group. Overall, the credit score model building submodule can build an accurate and effective personal credit score model that meets actual needs through the collaborative work of these submodules, providing strong technical support for accurate assessment of personal credit, and helping to improve the performance and application value of the credit enhancement assessment system.

[0167] Embodiment 8:

[0168] Based on Example 1, the second credit enhancement assessment execution module includes:

[0169] The family member credit index information acquisition submodule is used to determine that the current individual needs credit enhancement assessment and obtain the credit-related index information of the current individual's family members if it is determined that the current individual's credit score is lower than the score threshold corresponding to the target bank's loan amount;

[0170] The high-quality index value screening submodule is used to determine the score value of each secondary index under each primary index for each family member based on the credit-related index information of the current individual's family members and the preset family credit index evaluation standard, and screen out the maximum score value from the score values ​​of each secondary index under each primary index of all family members as the high-quality index value of the current individual's family members under the corresponding secondary index under the corresponding primary index;

[0171] A family credit score calculation submodule is used to calculate the family credit score of the current individual based on the high-quality index values ​​of each secondary index under each primary index of the family members of the current individual;

[0172] The family credit enhancement score calculation submodule is used to calculate the current individual's credit score and the family credit score according to the weight, and obtain the current individual's family credit enhancement score as the current individual's credit enhancement assessment result.

[0173] In this embodiment, the credit-related indicator information of family members refers to various specific data and situations of the family members in terms of credit, such as occupational status, financial status, credit record, etc.

[0174] In this embodiment, the preset family credit index evaluation standard is a set of rules and standards set in advance for evaluating the family credit status, including specific requirements and scoring methods for each indicator.

[0175] In this embodiment, based on the credit-related indicator information of the current individual's family members and the preset family credit indicator evaluation standard, the score value of each secondary indicator of each family member under each primary indicator is determined as follows:

[0176] Based on the actual credit-related situation of the current individual family members and in accordance with pre-set evaluation criteria, the specific scores of each family member for secondary indicators such as health status, property status, etc. under primary indicators such as basic situation and financial status are calculated.

[0177] The beneficial effects of the above technology are as follows: when the personal credit score value does not meet the standard, the family member credit index information acquisition submodule promptly obtains the credit-related index information of the family members to provide data support for subsequent evaluation. The high-quality index value screening submodule selects high-quality index values ​​according to the preset standards, ensuring the high quality and representativeness of the selected indicators. The family credit score value calculation submodule accurately calculates the family credit score value based on the high-quality index value, thereby improving the accuracy of the evaluation. The family credit enhancement score value calculation submodule calculates the personal credit score value and the family credit score value by weight to obtain the family credit enhancement score value, taking into account personal and family factors. On the whole, the second credit enhancement assessment execution module can comprehensively and accurately assess individuals who need credit enhancement assessment through the collaborative work of these submodules, provide a more reliable basis for banks' lending decisions, and help reduce credit risks and optimize credit resource allocation.

[0178] Embodiment 9:

[0179] On the basis of Example 8, the method for the family credit score calculation submodule to calculate the family credit score of the current individual based on the high-quality index value of each secondary index under each primary index of the family members of the current individual includes:

[0180]

[0181] In the formula, Z k is the credit score value of the kth primary indicator, q k is the total number of all secondary indicators under the kth primary indicator, Z k,x is the quality index value of the current individual’s family member under the xth secondary index under the kth primary index, W k,x is the preset weight of the xth secondary indicator under the kth primary indicator, M is the current individual's family credit score, p is the total number of all primary indicators, W k is the preset weight of the kth first-level indicator.

[0182] In this embodiment, the preset weight of the secondary indicator is a numerical value of relative importance set in advance for each secondary indicator when evaluating family credit. It is used to determine the influence of the secondary indicator score on the final result when calculating the total score.

[0183] In this embodiment, the preset weight of the first-level indicator is a numerical value set in advance for each first-level indicator (such as basic situation, financial situation, etc.) to indicate its importance in the comprehensive evaluation. This weight determines the influence of the first-level indicator in the overall evaluation.

[0184] In this embodiment, the preset weights of the secondary indicators and the preset weights of the primary indicators can generally be determined by the following methods:

[0185] Expert evaluation method: Invite experts in the field of credit assessment to make subjective judgments on the importance of each indicator based on their experience and expertise, thereby determining the weights;

[0186] Data analysis method: By analyzing a large amount of historical credit data, we observe the correlation and influence of different indicators on the final credit results, and determine the weights;

[0187] Analytic Hierarchy Process (AHP): decompose the problem into different levels, build a judgment matrix by comparing the relative importance of each indicator two by two, and calculate the weight;

[0188] Principal component analysis: conduct a comprehensive analysis of multiple related indicators, extract the main components, and determine the weights based on the degree of explanation of the main components to the original variables;

[0189] Goal-oriented method: According to the specific goals and purposes of credit assessment, determine which indicators are more critical to achieving the goals and assign corresponding weights.

[0190] The beneficial effects of the above technology are: the calculation formula clarifies the weight and contribution of each primary indicator and secondary indicator in the family credit score, making the score calculation more scientific and accurate. By setting weights for primary indicators and secondary indicators respectively, the importance of credit performance of family members in different aspects can be more carefully reflected. This calculation method comprehensively considers credit indicators in multiple dimensions, comprehensively evaluates the family's credit status, and improves the accuracy and reliability of family credit scores. It provides a quantitative calculation basis for the second credit enhancement assessment execution module, helps to more objectively judge the individual's family credit enhancement situation, and provides strong support for the bank's lending decisions. Overall, this calculation method improves the scientificity and effectiveness of the credit enhancement assessment system, can more accurately reflect the family's credit level, and reduce the error and risk of credit assessment.

[0191] Embodiment 10:

[0192] On the basis of Example 8, the court credit enhancement score value calculation submodule includes:

[0193] The family credit enhancement score calculation unit is used to calculate the current individual credit score and the family credit score according to the weight to obtain the current individual family credit enhancement score:

[0194] F=PW p +MW m

[0195] In the formula, F is the current individual's family credit enhancement score, P is the current individual's credit score, and W p is the preset weight of the current individual's credit score, M is the current individual's family credit score, and Wm The preset weight for the current individual's household credit score value;

[0196] The credit enhancement evaluation result output unit is used to use the current individual's family credit enhancement score value as the current individual's credit enhancement evaluation result.

[0197] In this embodiment, the preset weight of the current individual's credit score value refers to a numerical value preset to indicate the relative importance of the current individual's own credit score value when calculating the current individual's family credit enhancement score value.

[0198] In this embodiment, the preset weight of the current individual's family credit score is a numerical value determined in advance during the process of calculating the family credit enhancement score to give the current individual a degree of importance of the family credit score based on the credit status of family members.

[0199] In this embodiment, the preset weight of the current individual credit score value and the preset weight of the current individual family credit score value can generally be determined by the following methods:

[0200] Expert evaluation method: Invite experts in the field of credit assessment to make subjective judgments on the importance of each indicator based on their experience and expertise, thereby determining the weights;

[0201] Data analysis method: By analyzing a large amount of historical credit data, we observe the correlation and influence of different indicators on the final credit results, and determine the weights;

[0202] Analytic Hierarchy Process (AHP): decompose the problem into different levels, build a judgment matrix by comparing the relative importance of each indicator two by two, and calculate the weight;

[0203] Principal component analysis: conduct a comprehensive analysis of multiple related indicators, extract the main components, and determine the weights based on the degree of explanation of the main components to the original variables;

[0204] Goal-oriented method: According to the specific goals and purposes of credit assessment, determine which indicators are more critical to achieving the goals and assign corresponding weights.

[0205] The beneficial effects of the above technology are as follows: the family credit enhancement score calculation unit calculates the personal credit score and the family credit score according to the preset weights, making the calculation of the family credit enhancement score more reasonable and accurate. The weight ratio of personal credit and family credit in the final score is clarified, and the impact of the two factors on the credit assessment is comprehensively considered. The credit enhancement assessment result output unit outputs the calculated family credit enhancement score as the credit enhancement assessment result, providing a clear basis for subsequent decision-making. On the whole, the family credit enhancement score calculation submodule can accurately and effectively calculate the family credit enhancement score, and use it as an assessment result, providing more comprehensive and objective credit assessment data for bank lending decisions, which helps to improve the scientificity and accuracy of credit decisions and reduce credit risks.

[0206] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A credit enhancement evaluation system based on knowledge graph and deep learning, characterized in that: include: Personal credit rating calculation module, used to calculate the current personal credit score based on knowledge graph and deep learning methods; The first credit enhancement evaluation execution module is used to determine that the current individual does not need credit enhancement evaluation if it is determined that the current individual's credit score is not lower than the score threshold corresponding to the target bank's loan limit, and use the current individual's credit score as the current individual's credit enhancement evaluation result; The second credit enhancement assessment execution module is used to determine that the current individual needs a credit enhancement assessment if it is determined that the current individual's credit score is lower than the score threshold corresponding to the target bank's loan amount, and to screen out the high-quality indicator values ​​of each secondary indicator under each primary indicator from the credit-related indicator information of the current individual's family members, and calculate the current individual's family credit score based on the high-quality indicator values ​​of all secondary indicators under all primary indicators, and calculate the current individual's credit score and family credit score by weight to obtain the current individual's family credit enhancement score as the current individual's credit enhancement assessment result.

2. The credit enhancement evaluation system based on knowledge graph and deep learning according to claim 1 is characterized in that: Also includes: The first credit approval module is used to input the current individual's credit score value into the loan process processing end if the current individual's credit enhancement assessment result is the current individual's credit score value; The second credit approval module is used to determine whether the current individual's family credit enhancement score is not lower than the score threshold corresponding to the target bank's loan amount if the current individual's credit enhancement assessment result is the current individual's family credit enhancement score. If so, the current individual's family credit enhancement score is input into the loan process processing end; otherwise, the current individual's family credit enhancement score is stored in the cloud.

3. The credit enhancement evaluation system based on knowledge graph and deep learning according to claim 1 is characterized in that: Personal credit rating calculation module, including: A credit-related information acquisition submodule, used to collect the credit-related information of the current individual and the credit-related information of multiple reference individuals from multiple data sources; A knowledge graph generation submodule, for generating a knowledge graph of the credit index items of the current individual and a knowledge graph of the credit index items of each reference individual based on the credit-related information of the current individual and the credit-related information of all reference individuals; The credit scoring model building submodule is used to build a personal credit scoring model based on the step-by-step loan repayment requirements and the knowledge graph of the original credit index items of all reference individuals; The personal credit scoring submodule is used to input the current individual's credit index item knowledge graph into the personal credit scoring model to obtain the current individual's credit score value.

4. The credit enhancement evaluation system based on knowledge graph and deep learning according to claim 3 is characterized in that: The knowledge graph generation submodule includes: A credit-related information interpretation unit, used to identify all valid credit index items of all personal credit-related information and determine the inference logic between all valid credit index items; A knowledge graph framework generation unit is used to treat each credit index item as an entity, and construct a knowledge graph framework for all credit index items based on the inference logic between all credit index items, so as to obtain the original credit index item knowledge graph framework of each individual; The entity attribute value normalization unit is used to normalize the performance value of the same entity in the original credit index item knowledge graph framework of all individuals in the credit-related information of all individuals, and obtain the normalized value of each entity in the credit-related information of each individual; A knowledge graph framework assignment unit, used to assign a value to the credit index item knowledge graph framework of each individual based on the normalized value of each entity in the credit-related information of each individual, so as to obtain the credit index item knowledge graph of each individual; Among them, the knowledge graph of credit index items of all individuals includes the knowledge graph of credit index items of the current individual and the knowledge graph of credit index items of all reference individuals.

5. The credit enhancement evaluation system based on knowledge graph and deep learning according to claim 4 is characterized in that: Credit related information interpretation unit, including: The inference logic generation subunit is used to determine all characteristic attributes of each piece of information in all personal credit-related information, trace back the numerical inference logic of the characteristic attributes of all pieces of information in each personal credit-related information, and obtain the inference logic between all characteristic attributes; A marginal feature attribute screening unit, used to treat the feature attributes that do not involve any inference logic among all the feature attributes of each individual's credit-related information as marginal feature attributes; A contribution rate calculation unit is used to calculate the contribution rate of all marginal feature attributes in all personal credit related information; The effective index item screening unit is used to treat all the marginal characteristic attributes whose contribution rate in all personal credit-related information is not less than the contribution rate threshold and all the characteristic attributes involved in the inference logic in the corresponding credit-related information as the effective credit index items in the corresponding credit-related information; An upper-layer inference logic evolution unit, used to perform upper-layer inference logic evolution on each valid credit index item not involving inference logic, and obtain an upper-layer inference logic evolution result of each valid credit index item not involving inference logic; The inference logic merging unit is used to merge the upper-level inference logic evolution results of all valid credit index items that do not involve the inference logic with the inference logic between all feature attributes to obtain the inference logic between all valid credit index items.

6. The credit enhancement evaluation system based on knowledge graph and deep learning according to claim 5 is characterized in that: The upper layer inference logic evolution unit includes: The co-occurrence count subunit is used to obtain a large amount of credit-related information, treat each valid credit index item not involving inference logic as a marginal valid credit index item, and count the co-occurrence times of each marginal valid credit index item and each valid credit index item involving inference logic in all credit-related information; A co-occurrence frequency calculation subunit, used to take the quotient of the sum of the number of characters of each marginal effective credit index item and each effective credit index item involved in the inference logic and the corresponding number of co-occurrences in all copies of credit-related information and the sum of the number of characters in all copies of credit-related information as the co-occurrence frequency of each marginal effective credit index item and each effective credit index item involved in the inference logic in all copies of credit-related information; A credit index item screening subunit is used to use all valid credit index items involved in the inference logic with a co-occurrence frequency of each marginal valid credit index item in all credit-related data that is not less than a preset co-occurrence frequency as the basis credit index items of each marginal valid credit index item; The upper-level inference logic evolution subunit is used to mine the inference logic between each marginal effective credit indicator item and all corresponding based credit indicator items as the upper-level inference logic evolution result of each marginal effective credit indicator item based on all co-occurring sentences between each marginal effective credit indicator item and all corresponding based credit indicator items in all credit-related materials.

7. The credit enhancement evaluation system based on knowledge graph and deep learning according to claim 3 is characterized in that: The credit scoring model building submodule includes: The excellent sample screening submodule is used to screen out excellent individual sample groups for each loan amount value in the ladder loan amount from all reference individuals based on the ladder loan repayment requirements; The loan quota and credit score value alignment submodule is used to align each loan quota value with the enterprise's loan quota credit score benchmark to obtain the standard credit score value corresponding to each loan quota value; The credit scoring model building submodule is used to conduct deep learning and model building on the standard credit scoring value corresponding to each loan limit value and the original credit index item knowledge graph of all reference individuals in the corresponding excellent individual sample group, so as to construct a personal credit scoring model.

8. The credit enhancement evaluation system based on knowledge graph and deep learning according to claim 1 is characterized in that: The second credit enhancement assessment execution module includes: The family member credit index information acquisition submodule is used to determine that the current individual needs credit enhancement assessment and obtain the credit-related index information of the current individual's family members if it is determined that the current individual's credit score is lower than the score threshold corresponding to the target bank's loan amount; The high-quality index value screening submodule is used to determine the score value of each secondary index under each primary index for each family member based on the credit-related index information of the current individual's family members and the preset family credit index evaluation standard, and screen out the maximum score value from the score values ​​of each secondary index under each primary index of all family members as the high-quality index value of the current individual's family members under the corresponding secondary index under the corresponding primary index; A family credit score calculation submodule is used to calculate the family credit score of the current individual based on the high-quality index values ​​of each secondary index under each primary index of the family members of the current individual; The family credit enhancement score calculation submodule is used to calculate the current individual's credit score and the family credit score according to the weight, and obtain the current individual's family credit enhancement score as the current individual's credit enhancement assessment result.

9. The credit enhancement evaluation system based on knowledge graph and deep learning according to claim 8, characterized in that: The method of the family credit score calculation submodule to calculate the family credit score of the current individual based on the high-quality index values ​​of each secondary index under each primary index of the family members of the current individual, includes: In the formula, Z k is the credit score value of the kth primary indicator, q k is the total number of all secondary indicators under the kth primary indicator, Z k,x is the quality index value of the current individual’s family member under the xth secondary index under the kth primary index, W k,x is the preset weight of the xth secondary indicator under the kth primary indicator, M is the current individual's family credit score, p is the total number of all primary indicators, W k is the preset weight of the kth first-level indicator.

10. The credit enhancement evaluation system based on knowledge graph and deep learning according to claim 8, characterized in that: The family credit enhancement score calculation submodule includes: The family credit enhancement score calculation unit is used to calculate the current individual credit score and the family credit score according to the weight to obtain the current individual family credit enhancement score: F=PW p +MW m In the formula, F is the current individual's family credit enhancement score, P is the current individual's credit score, and W p is the preset weight of the current individual's credit score, M is the current individual's family credit score, and W m The preset weight for the current individual's household credit score value; The credit enhancement evaluation result output unit is used to use the current individual's family credit enhancement score value as the current individual's credit enhancement evaluation result.

Images