Block chain finance enterprise credit evaluation system and method

Abstract

The invention relates to the technical field of enterprise credit evaluation, and discloses a blockchain finance enterprise credit evaluation system, which comprises an enterprise credit evaluation system, and the enterprise credit evaluation system comprises a multi-source data acquisition layer, a credit processing engine layer, an algorithm and model layer, an intelligent contract layer and an application service layer. The multi-source data acquisition layer comprises a multi-dimensional data unit, an enhanced data management and control unit and an external data introduction unit; the credit processing engine layer comprises a data cleaning and verification unit, a learning model training unit and an association network analysis unit. According to the enterprise credit assessment system and method for block chain finance, the multi-source data acquisition layer, the credit processing engine layer, the algorithm and model layer, the intelligent contract layer and the application service layer are arranged, layered architecture design is adopted, bottom-layer data acquisition is performed, upper-layer application service is performed, all the layers are clear in division of labor, and an enterprise credit assessment function is cooperatively realized; the system has the characteristics of clear modularization and strong expansibility.

Description

Technical Field

[0001] The present invention relates to the technical field of enterprise credit assessment, and in particular to a blockchain finance enterprise credit assessment system and method. Background Art

[0002] Financial enterprises are mainly engaged in deposit and lending, financing and mergers and acquisitions, insurance, securities and funds, etc. Their credit is the "credit card" of financial enterprises in market economic activities and the foundation of engaging in financial business. Therefore, accurate assessment of the credit of financial enterprises is of vital importance to financial enterprises. Enterprise credit assessment is an assessment of the performance of enterprises in various aspects of economic and social activities, such as compliance with regulatory laws and regulations, fulfillment of various economic contracts, and organization and management of their behavior. It emphasizes the evaluation of the credit of enterprises in various aspects, such as compliance with national laws, regulations and policies, fulfillment of corporate commitments and contracts, and the quality of corporate organization and management, and focuses on evaluating the past performance of enterprises.

[0003] Currently, the assessment of financial enterprise credit information in the market relies more on the subjective judgment of loan officers, who assess risks by combining information such as the enterprise's credit history, financial data, and collateral. However, with the advancement of big data and artificial intelligence technologies, corporate credit assessment is increasingly tending towards a data-driven decision-making model. This shift has greatly improved the efficiency of the loan approval process and enhanced the accuracy of risk control. At the same time, it is also prone to data security issues. Therefore, how to accurately, efficiently, and securely conduct credit assessments on financial enterprises is one of the technical issues that need to be addressed. Summary of the Invention

[0004] (1) Technical problems solved

[0005] In response to the shortcomings of the existing technology, the present invention provides a blockchain finance enterprise credit assessment system and method, which has the advantages of clear modularization and strong scalability. It solves the problem that the assessment of existing financial enterprise credit information relies more on the subjective judgment of loan officers. It combines the enterprise's credit history, financial data, collateral and other information to assess risks. However, with the advancement of big data and artificial intelligence technology, enterprise credit assessment is increasingly tending towards a data-driven decision-making model. This transformation greatly improves the efficiency of the loan approval process and enhances the accuracy of risk control. At the same time, it is also prone to data security issues.

[0006] (2) Technical solution

[0007] To achieve the above-mentioned goals of clear modularization and strong scalability, the present invention provides the following technical solutions: a blockchain finance enterprise credit assessment system and method, including an enterprise credit assessment system, wherein the enterprise credit assessment system includes a multi-source data acquisition layer, a credit processing engine layer, an algorithm and model layer, a smart contract layer, and an application service layer;

[0008] The multi-source data acquisition layer includes a multi-dimensional data unit, an enhanced data management unit, and an external data introduction unit;

[0009] The credit processing engine layer includes a data cleaning and verification unit, a learning model training unit, and an association network analysis unit;

[0010] The algorithm and model layer includes optimization of credit assessment model, model integration and fusion, and model integration and fusion;

[0011] The smart contract layer includes a smart contract optimization unit, a privacy computing upgrade unit, and a blockchain technology improvement unit;

[0012] The application service layer provides specific application services to enterprises, financial institutions, regulatory authorities, etc. by integrating the results of processing and analysis by the previous layers.

[0013] Furthermore, the multi-dimensional data unit connects to data sources such as the enterprise's internal system (ERP, etc.), blockchain network, and IoT devices, and captures data regularly or in real time through API interfaces or data transmission protocols. The enhanced data management and control unit uses data quality assessment indicators and algorithms, such as detecting missing values ​​and outliers in data, and uses data cleaning algorithms (such as data smoothing and deduplication) to process the original data. The external data introduction unit cooperates with external data providers to obtain external data regularly or on demand and integrate them through purchasing data services, data sharing agreements, etc.

[0014] Furthermore, the data cleaning and verification unit performs secondary screening and correction on the data by utilizing data verification rules (such as format verification, value range verification, etc.) and cleaning algorithms. The learning model training unit uses machine learning, deep learning and other algorithms, combined with collected and processed data, to train the credit assessment model and explore the correlation between data features and credit status.

[0015] Furthermore, the association network analysis unit constructs an enterprise association relationship graph and uses a graph algorithm (such as PageRank algorithm, etc.) to analyze the influence and association closeness between nodes (enterprises).

[0016] Furthermore, the optimized credit assessment model is used to compare the performance indicators of different models (such as accuracy, recall rate, F1 value, etc.), and model tuning techniques (such as hyperparameter adjustment, feature engineering optimization, etc.) are used to improve the model. The model integration and fusion adopts integrated learning methods (such as voting method, stacking method, bagging, boosting, etc.) to combine the prediction results of multiple base models to obtain the final evaluation conclusion.

[0017] Furthermore, the real-time model update is triggered by setting a model update mechanism, such as retraining the model and updating the parameters when the new data reaches a certain amount or at a fixed interval.

[0018] Furthermore, the smart contract optimization unit uses code review tools and best practice specifications to check the contract code logic, optimize the code structure, and adopt a secure programming paradigm. The privacy computing upgrade unit selects appropriate privacy computing technology based on business scenarios and data characteristics, encrypts the data involved in the contract, and performs calculations and verifications in a ciphertext state.

[0019] Furthermore, the blockchain technology improvement unit studies and compares the advantages and disadvantages of different consensus algorithms (such as PoW, PoS, DPoS, etc.), selects or improves the consensus algorithm based on system requirements, and optimizes the blockchain node configuration and network communication mechanism.

[0020] A method for enterprise credit assessment in blockchain finance, including the above-mentioned enterprise credit assessment system for blockchain finance, has the following operating steps:

[0021] Step S1: Through the multi-dimensional data units of the multi-source data collection layer, enterprise-related data is collected from different channels such as on-chain (transactions / contracts), off-chain (ERP / tax), and IoT real-time data streams to comprehensively collect information on enterprise operations and transactions;

[0022] Step S2: Strengthen the data management and control unit to control the quality of the collected raw data, check the accuracy, completeness and consistency of the data, clean the data, remove noise and duplicate values, etc. At the same time, introduce external data units to obtain external data such as macroeconomics, industry, and government public data to enrich the evaluation data dimensions;

[0023] Step S3: At the credit processing engine layer, the data cleaning and verification unit performs a further in-depth cleaning and verification of the previously collected and preliminarily processed data to ensure that the data quality meets the requirements;

[0024] Step S4: Using machine learning, deep learning and other algorithms through the learning model training unit, the processed data is divided into training sets and test sets, the credit assessment model is trained, and the correlation between data features and the credit status of the enterprise is mined;

[0025] Step S5: Use the association network analysis unit to build an association network between the enterprise and its upstream and downstream partners, and use graph algorithms to analyze the credit transmission and risk diffusion of the enterprise in the network;

[0026] Step S6: Continuously improve the credit assessment model by optimizing the credit assessment model unit, comparing the performance indicators of different models, and using hyperparameter adjustment, feature engineering optimization and other technologies to improve model accuracy.

[0027] Step S7: Using ensemble learning methods such as voting and stacking methods through the model integration and fusion unit, the results of multiple different types of credit assessment models are combined to reduce the bias and risk of a single model;

[0028] Step S8: Setting a model update trigger mechanism through the real-time model update unit, using the new data or time interval to retrain the model and update the parameters to adapt to the dynamic changes of the enterprise and the market;

[0029] Step S9: The smart contract optimization unit uses code review tools and secure programming paradigms to review and optimize the smart contract code to improve execution efficiency and enhance security.

[0030] Step S10: The privacy computing upgrade unit uses privacy computing technologies such as homomorphic encryption and secure multi-party computing to encrypt the data involved in the contract and perform calculations and verifications while ensuring data privacy.

[0031] Step S11: Through the blockchain technology improvement unit, research and select a more efficient consensus algorithm, optimize the blockchain network architecture and node configuration, improve the performance and scalability of the blockchain platform, and support the stable operation of smart contracts;

[0032] Step S12: The application service layer integrates the processing and analysis results of the previous stages to provide enterprises, financial institutions, regulatory authorities, etc. with application services such as dynamic credit score dashboards (real-time updates of enterprise credit scores), risk warning systems (smart contracts automatically trigger risk warnings), and credit token trading markets (realizing credit asset transactions).

[0033] (3) Beneficial effects

[0034] Compared with the existing technology, the present invention provides a blockchain finance enterprise credit assessment system and method, which has the following beneficial effects:

[0035] 1. This blockchain finance enterprise credit assessment system and method sets up a multi-source data collection layer, a credit processing engine layer, an algorithm and model layer, a smart contract layer, and an application service layer. By adopting a layered architecture design, from the bottom-level data collection to the upper-level application service, each layer has a clear division of labor and collaborates to achieve the enterprise credit assessment function. It has the characteristics of clear modularity and strong scalability.

[0036] 2. The enterprise credit assessment system and method of blockchain finance breaks through the limitations of traditional single data sources through multi-dimensional data units, and widely collects multi-channel data such as on-chain, off-chain and IoT real-time data streams to achieve strong data comprehensiveness. It can portray the enterprise portrait from multiple perspectives such as enterprise operations, transactions, and IoT perception, reduce information asymmetry, and improve assessment accuracy. By strengthening the data management and control unit, it can pay attention to data quality, strictly control and pre-process the data source, and through active data quality detection and cleaning, ensure the reliability of data entering subsequent links, lay the foundation for accurate assessment, and reduce assessment deviations caused by data problems. By introducing external data units, it can introduce external data such as macroeconomics, industry, and government public, expand the boundaries of assessment data, enable the assessment to combine external environmental factors, more objectively reflect the credit status of the enterprise in the industry and macroeconomic context, and enhance the scientific nature and foresight of the assessment.

[0037] 3. The enterprise credit assessment system and method of blockchain finance uses a data cleaning and verification unit to perform secondary in-depth processing on the data. Different from the general simple cleaning after collection, it further improves the data quality, ensures the accuracy of the data input into the model, reduces the interference of noise and erroneous data on the assessment results, and improves the reliability of the assessment. Through the learning model training unit, it can use advanced algorithms such as machine learning and deep learning to train the credit assessment model, which can explore the deep characteristics and complex relationships of the data and adapt to the nonlinear and dynamic changes in the enterprise credit assessment. Compared with traditional models, it has higher accuracy and generalization ability. Through the association network analysis unit, it can analyze credit transmission and risk diffusion from the perspective of the enterprise association network, breaking through the limitations of isolated assessment of enterprise credit, considering the associated impact of enterprises in the business ecosystem, more comprehensively assessing credit risks, and discovering potential risk transmission paths.

[0038] 4. The enterprise credit assessment system and method of blockchain finance can focus on optimizing smart contract codes through the smart contract optimization unit, improve execution efficiency and security, ensure efficient and stable operation of smart contracts in credit assessment applications, reduce contract vulnerabilities and security risks, and enhance system reliability. Through the privacy computing upgrade unit, advanced privacy computing technology can be used to protect data privacy, ensure that sensitive enterprise data is not leaked during data sharing and calculation, balance data utilization and privacy protection, promote data compliance circulation and multi-party collaboration, and carry out targeted improvements and optimizations on the underlying blockchain technology through the blockchain technology improvement unit to improve the performance and scalability of the blockchain platform, better support the operation of smart contracts, and ensure the safe and efficient data storage, transmission and processing of the credit assessment system. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the enterprise credit evaluation system of the present invention;

[0040] Figure 2This is a schematic diagram of the multi-source data acquisition layer of the present invention;

[0041] Figure 3 This is a schematic diagram of the credit processing engine layer of the present invention;

[0042] Figure 4 Schematic diagram of the algorithm and model layer of the present invention;

[0043] Figure 5 This is a schematic diagram of the smart contract layer of the present invention;

[0044] Figure 6 This is a flow chart of the evaluation system of the present invention. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] See also Figure 1-6 , a blockchain finance enterprise credit assessment system, including an enterprise credit assessment system, the enterprise credit assessment system includes a multi-source data acquisition layer, a credit processing engine layer, an algorithm and model layer, a smart contract layer and an application service layer;

[0047] The multi-source data collection layer includes multi-dimensional data units, enhanced data management units, and external data introduction units;

[0048] The credit processing engine layer includes a data cleaning and verification unit, a learning model training unit, and an association network analysis unit;

[0049] The algorithm and model layer includes optimization of credit assessment models, model integration and fusion, and model integration and fusion;

[0050] The smart contract layer includes a smart contract optimization unit, a privacy computing upgrade unit, and a blockchain technology improvement unit;

[0051] The application service layer provides specific application services to enterprises, financial institutions, regulatory authorities, etc. by integrating the results of processing and analysis by the previous layers.

[0052] In the case implementation, the multi-dimensional data unit connects to data sources such as the enterprise's internal systems (ERP, etc.), blockchain networks, and IoT devices, and captures data regularly or in real time through API interfaces or data transmission protocols. The data management unit is strengthened to use data quality assessment indicators and algorithms, such as detecting missing values ​​and outliers, and using data cleaning algorithms (such as data smoothing and deduplication) to process raw data. The external data unit is introduced to cooperate with external data providers, and external data is obtained and integrated regularly or on demand through purchasing data services and data sharing agreements.

[0053] Among them, the calculation of data missing rate in the strengthened data control unit is used to measure the degree of data missing. The formula is:

[0054] Missing rate = (number of missing values / total number of data) × 100%;

[0055] For example, count the missing fields in corporate transaction data to determine whether it affects subsequent evaluations.

[0056] Strengthening data outlier detection in data control units - Z-score formula:

[0057] Z i =xi-xˉ / s, where x i is the i-th data point, xˉ is the mean, and s is the standard deviation. We use the Z-score to determine whether a data point is an outlier (e.g., |Z| > 3 is considered an outlier) and detect abnormal fluctuations in corporate financial data.

[0058] In the case implementation, the data cleaning and verification unit uses data verification rules (such as format verification and value range verification) and cleaning algorithms to conduct secondary screening and correction of data. The learning model training unit uses machine learning, deep learning and other algorithms, combined with collected and processed data, to train the credit assessment model and explore the correlation between data features and credit status. The association network analysis unit constructs an enterprise association relationship graph and uses graph algorithms (such as the PageRank algorithm) to analyze the influence and degree of correlation between nodes (enterprises).

[0059] Among them, the logistic regression in the learning model training unit is: For the two-class problem, let the input feature vector x = (x1, x2, ..., x n ), weight vector w=(w1,w2,…,w n ), the bias is b, then the prediction function is y = 1 / 1 + e -(w·x+b) , the y value represents the probability that the sample belongs to the positive class, based on which the corporate credit risk (such as default probability) is judged.

[0060] Decision tree-information gain calculation in the learning model training unit: When selecting the decision tree splitting node, the information gain is often used to measure information gain, IG(S, A) = H(S)-∑ v∈Values(A) ∣Sv∣ / ∣S∣H(S v ), where S is the sample set, A is the attribute, Values(A) is the value set of attribute A, S v is the sample subset when the attribute A takes the value v, H(S) is the information entropy of the sample set S H(S) = -∑ i c =1p i log2p i (c is the number of categories, p i is the proportion of samples in the i-th category), and the optimal splitting attribute is selected through information gain to construct a decision tree to evaluate corporate credit.

[0061] The association network analysis unit uses the PageRank algorithm: let the web page (enterprise node) set be V, the PageRank value of node i be PR(i), M(i) is the node set pointing to node i, L(j) is the number of links from node j to other nodes, and the damping coefficient d (generally 0.8-0.9), then the iterative formula is PR(i) = (1-d) / N+d∑ j∈M(i)PR(j) / L(j) (N is the total number of nodes), which is used to measure the importance of an enterprise in the association network.

[0062] The shortest path in the association network analysis unit - Dijkstra algorithm: used to calculate the shortest path from a specific source node to other nodes in the graph. Suppose the graph G = (V, E), the source node is s, the edge weight from node u to node v is w(u, v), and the distance array dist records the shortest distance from the source node to each node. Initially, dist[s] = 0, and the dist values ​​of other nodes are infinite. Each time, the node u with the smallest distance is selected from the nodes whose shortest path has not yet been determined, and the relaxation operation dist[v] = min(dist[v], dist[u] + w(u, v)) is performed on its adjacent node v. It is continuously iterated until the shortest paths of all nodes are determined. It can be used to analyze the transmission path of corporate credit risk in the association network.

[0063] In the case implementation, the credit assessment model was optimized to compare the performance indicators of different models (such as accuracy, recall, and F1 value). Model tuning techniques (such as hyperparameter adjustment and feature engineering optimization) were used to improve the model. Model integration and fusion used ensemble learning methods (such as voting, stacking, bagging, and boosting) to combine the prediction results of multiple base models to reach the final evaluation conclusion. Real-time model updates were triggered by setting a model update trigger mechanism. For example, when the amount of new data reached a certain level or at a fixed interval, the model was retrained and the parameters were updated.

[0064] Among them, the voting method of model integration and fusion: for classification problems, assume that there are k base models, each model predicts the category of sample x as yi(x) (i=1, 2,…, k), and the final prediction category y(x) is determined by the prediction category of the majority of base models. For example, k=3 models predict the corporate credit rating as "high", "medium" and "high" respectively, then the final prediction is "high".

[0065] Real-time model update in model integration and fusion - gradient descent method (for model parameter update): Assume the loss function L(θ), θ is the model parameter vector, and the learning rate is α, then the parameter update formula is in It is the gradient of the loss function with respect to the parameter θ. By continuously iteratively updating the parameters, the model can adapt to new data and improve the accuracy of credit assessment.

[0066] During the case implementation, the smart contract optimization unit used code review tools and best practice specifications to check the contract code logic, optimize the code structure, and adopt a secure programming paradigm. The privacy computing upgrade unit selected appropriate privacy computing technology based on business scenarios and data characteristics, encrypting the data involved in the contract, and performing calculations and verifications in ciphertext. The blockchain technology improvement unit studied and compared the advantages and disadvantages of different consensus algorithms (such as PoW, PoS, DPoS, etc.), selected or improved the consensus algorithm based on system requirements, and optimized the blockchain node configuration and network communication mechanism.

[0067] A method for enterprise credit assessment in blockchain finance, including the above-mentioned enterprise credit assessment system for blockchain finance, has the following operating steps:

[0068] Step S1: Through the multi-dimensional data units of the multi-source data collection layer, enterprise-related data is collected from different channels such as on-chain (transactions / contracts), off-chain (ERP / tax), and IoT real-time data streams to comprehensively collect information on enterprise operations and transactions;

[0069] Step S2: Strengthen the data management and control unit to control the quality of the collected raw data, check the accuracy, completeness and consistency of the data, clean the data, remove noise and duplicate values, etc. At the same time, introduce external data units to obtain external data such as macroeconomics, industry, and government public data to enrich the evaluation data dimensions;

[0070] Step S3: At the credit processing engine layer, the data cleaning and verification unit performs a further in-depth cleaning and verification of the previously collected and preliminarily processed data to ensure that the data quality meets the requirements;

[0071] Step S4: Using machine learning, deep learning and other algorithms through the learning model training unit, the processed data is divided into training sets and test sets, the credit assessment model is trained, and the correlation between data features and the credit status of the enterprise is mined;

[0072] Step S5: Use the association network analysis unit to build an association network between the enterprise and its upstream and downstream partners, and use graph algorithms to analyze the credit transmission and risk diffusion of the enterprise in the network;

[0073] Step S6: Continuously improve the credit assessment model by optimizing the credit assessment model unit, comparing the performance indicators of different models, and using hyperparameter adjustment, feature engineering optimization and other technologies to improve model accuracy.

[0074] Step S7: Using ensemble learning methods such as voting and stacking methods through the model integration and fusion unit, the results of multiple different types of credit assessment models are combined to reduce the bias and risk of a single model;

[0075] Step S8: Setting a model update trigger mechanism through the real-time model update unit, using the new data or time interval to retrain the model and update the parameters to adapt to the dynamic changes of the enterprise and the market;

[0076] Step S9: The smart contract optimization unit uses code review tools and secure programming paradigms to review and optimize the smart contract code to improve execution efficiency and enhance security.

[0077] Step S10: The privacy computing upgrade unit uses privacy computing technologies such as homomorphic encryption and secure multi-party computing to encrypt the data involved in the contract and perform calculations and verifications while ensuring data privacy.

[0078] Step S11: Through the blockchain technology improvement unit, research and select a more efficient consensus algorithm, optimize the blockchain network architecture and node configuration, improve the performance and scalability of the blockchain platform, and support the stable operation of smart contracts;

[0079] Step S12: The application service layer integrates the processing and analysis results of the previous stages to provide enterprises, financial institutions, regulatory authorities, etc. with application services such as dynamic credit score dashboards (real-time updates of enterprise credit scores), risk warning systems (smart contracts automatically trigger risk warnings), and credit token trading markets (realizing credit asset transactions).

[0080] In summary, the enterprise credit assessment system and method for blockchain finance, by setting up a multi-source data collection layer, a credit processing engine layer, an algorithm and model layer, a smart contract layer, and an application service layer, adopts a layered architecture design. From bottom-level data collection to upper-level application services, each layer has a clear division of labor and collaborates to achieve the enterprise credit assessment function. It has the characteristics of clear modularity and strong scalability. Through multi-dimensional data units, it breaks through the limitations of traditional single data sources and widely collects multi-channel data such as on-chain, off-chain, and IoT real-time data streams to achieve strong data comprehensiveness. It can portray enterprise portraits from multiple perspectives such as enterprise operations, transactions, and IoT perception, reduce information asymmetry, and improve assessment accuracy. By strengthening the data management unit, it can pay attention to data quality, strictly control and preprocess the data source, and through active data quality detection and cleaning, ensure the reliability of data entering subsequent links, lay the foundation for accurate assessment, and reduce assessment bias caused by data problems. By introducing external data units, it can introduce external data such as macroeconomic, industry, and government public data, expand the boundaries of assessment data, enable the assessment to combine external environmental factors, more objectively reflect the credit status of enterprises in the industry and macroeconomic context, and enhance the scientific nature and foresight of the assessment.

[0081] In addition, the data is processed in depth for the second time through the data cleaning and verification unit, which is different from the general simple cleaning after collection, further improving the data quality, ensuring the accuracy of the data input to the model, reducing the interference of noise and erroneous data on the evaluation results, and improving the reliability of the evaluation. The credit evaluation model can be trained by the learning model training unit using advanced algorithms such as machine learning and deep learning, which can mine the deep characteristics and complex relationships of the data and adapt to the nonlinear and dynamic changes in the corporate credit evaluation. Compared with the traditional model, it has higher accuracy and generalization ability. The association network analysis unit can analyze the credit transmission and risk diffusion from the perspective of the corporate association network, breaking through the limitations of isolated corporate credit evaluation and considering the impact of the correlation of enterprises in the business ecosystem. Assess credit risk comprehensively and discover potential risk transmission paths. Through the smart contract optimization unit, you can focus on optimizing smart contract codes, improve execution efficiency and security, ensure efficient and stable operation of smart contracts in credit assessment applications, reduce contract vulnerabilities and security risks, and enhance system reliability. Through the privacy computing upgrade unit, you can use advanced privacy computing technology to protect data privacy, ensure that sensitive corporate data is not leaked during data sharing and calculation, balance data utilization and privacy protection, promote data compliance circulation and multi-party collaboration, and through the blockchain technology improvement unit, make targeted improvements and optimizations to the underlying blockchain technology, improve the performance and scalability of the blockchain platform, better support the operation of smart contracts, and ensure the safe and efficient storage, transmission and processing of data in the credit assessment system.

[0082] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0083] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A blockchain finance enterprise credit assessment system, including an enterprise credit assessment system, characterized by: The enterprise credit assessment system includes a multi-source data collection layer, a credit processing engine layer, an algorithm and model layer, a smart contract layer, and an application service layer; The multi-source data acquisition layer includes a multi-dimensional data unit, an enhanced data management unit, and an external data introduction unit; The credit processing engine layer includes a data cleaning and verification unit, a learning model training unit, and an association network analysis unit; The algorithm and model layer includes optimization of credit assessment model, model integration and fusion, and model integration and fusion; The smart contract layer includes a smart contract optimization unit, a privacy computing upgrade unit, and a blockchain technology improvement unit; The application service layer provides specific application services to enterprises, financial institutions, regulatory authorities, etc. by integrating the results of processing and analysis by the previous layers.

2. The enterprise credit assessment system for blockchain finance according to claim 1, characterized in that: The multi-dimensional data unit connects to data sources such as the enterprise's internal system (ERP, etc.), blockchain network, and IoT devices, and captures data regularly or in real time through API interfaces or data transmission protocols. The enhanced data management and control unit uses data quality assessment indicators and algorithms, such as detecting missing values ​​and outliers in data, and uses data cleaning algorithms (such as data smoothing and deduplication) to process the original data. The external data introduction unit cooperates with external data providers to obtain external data regularly or on demand and integrate them through purchasing data services, data sharing agreements, etc.

3. The blockchain finance enterprise credit assessment system according to claim 1, characterized in that: The data cleaning and verification unit performs secondary screening and correction on the data by using data verification rules (such as format verification, value range verification, etc.) and cleaning algorithms. The learning model training unit uses machine learning, deep learning and other algorithms, combined with collected and processed data, to train the credit assessment model and explore the correlation between data features and credit status.

4. The blockchain finance enterprise credit assessment system according to claim 1, characterized in that: The association network analysis unit constructs an enterprise association relationship graph and uses a graph algorithm (such as PageRank algorithm, etc.) to analyze the influence and association closeness between nodes (enterprises).

5. The blockchain finance enterprise credit assessment system according to claim 1, characterized in that: The optimized credit assessment model is used to compare the performance indicators of different models (such as accuracy, recall rate, F1 value, etc.), and improve the model using model tuning techniques (such as hyperparameter adjustment, feature engineering optimization, etc.). The model integration and fusion adopts integrated learning methods (such as voting method, stacking method, bagging, boosting, etc.) to combine the prediction results of multiple base models to obtain the final evaluation conclusion.

6. The blockchain finance enterprise credit assessment system according to claim 1, characterized in that: The real-time model update is achieved by setting a model update trigger mechanism, such as retraining the model and updating the parameters when the new data reaches a certain amount or at a fixed interval.

7. The blockchain finance enterprise credit assessment system according to claim 1, characterized in that: The smart contract optimization unit uses code review tools and best practice specifications to check the contract code logic, optimize the code structure, and adopt a secure programming paradigm. The privacy computing upgrade unit selects appropriate privacy computing technology based on business scenarios and data characteristics, encrypts the data involved in the contract, and performs calculations and verifications in a ciphertext state.

8. The blockchain finance enterprise credit assessment system according to claim 1, characterized in that: The blockchain technology improvement unit studies and compares the advantages and disadvantages of different consensus algorithms (such as PoW, PoS, DPoS, etc.), selects or improves the consensus algorithm based on system requirements, and optimizes the blockchain node configuration and network communication mechanism.

9. A method for enterprise credit assessment in blockchain finance, comprising the enterprise credit assessment system for blockchain finance according to claims 1-8, characterized in that: The steps are as follows: Step S1: Through the multi-dimensional data units of the multi-source data collection layer, enterprise-related data is collected from different channels such as on-chain (transactions / contracts), off-chain (ERP / tax), and IoT real-time data streams to comprehensively collect information on enterprise operations and transactions; Step S2: Strengthen the data management and control unit to control the quality of the collected raw data, check the accuracy, completeness and consistency of the data, clean the data, remove noise and duplicate values, etc. At the same time, introduce external data units to obtain external data such as macroeconomics, industry, and government public data to enrich the evaluation data dimensions; Step S3: At the credit processing engine layer, the data cleaning and verification unit performs a further in-depth cleaning and verification of the previously collected and preliminarily processed data to ensure that the data quality meets the requirements; Step S4: Using machine learning, deep learning and other algorithms through the learning model training unit, the processed data is divided into training sets and test sets, the credit assessment model is trained, and the correlation between data features and the credit status of the enterprise is mined; Step S5: Use the association network analysis unit to build an association network between the enterprise and its upstream and downstream partners, and use graph algorithms to analyze the credit transmission and risk diffusion of the enterprise in the network; Step S6: Continuously improve the credit assessment model by optimizing the credit assessment model unit, comparing the performance indicators of different models, and using hyperparameter adjustment, feature engineering optimization and other technologies to improve model accuracy. Step S7: Using ensemble learning methods such as voting and stacking methods through the model integration and fusion unit, the results of multiple different types of credit assessment models are combined to reduce the bias and risk of a single model; Step S8: Setting a model update trigger mechanism through the real-time model update unit, using the new data or time interval to retrain the model and update the parameters to adapt to the dynamic changes of the enterprise and the market; Step S9: The smart contract optimization unit uses code review tools and secure programming paradigms to review and optimize the smart contract code to improve execution efficiency and enhance security. Step S10: The privacy computing upgrade unit uses privacy computing technologies such as homomorphic encryption and secure multi-party computing to encrypt the data involved in the contract and perform calculations and verifications while ensuring data privacy. Step S11: Through the blockchain technology improvement unit, research and select a more efficient consensus algorithm, optimize the blockchain network architecture and node configuration, improve the performance and scalability of the blockchain platform, and support the stable operation of smart contracts; Step S12: The application service layer integrates the processing and analysis results of the previous stages to provide enterprises, financial institutions, regulatory authorities, etc. with application services such as dynamic credit score dashboards (real-time updates of enterprise credit scores), risk warning systems (smart contracts automatically trigger risk warnings), and credit token trading markets (realizing credit asset transactions).

Images