On-chain enterprise risk identification method and system based on credit chain large model

Abstract

The invention discloses an on-chain enterprise risk identification method and system based on a credit chain large model. The system comprises a data preprocessing module, a model matching module, a weight parameter loading module, a data operation reasoning module, a risk prediction output module and the like. Large model training and optimization are carried out by integrating industry, region, type and other multi-dimensional data sets, a credit chain vertical type general model and a special model are formed, and the method can adapt to credit and risk prediction evaluation, monitoring and early warning of a single enterprise or an integrated enterprise in different supply chain and industry chain scenes. A more comprehensive, accurate and real-time solution is provided for enterprise management and industry supervision, and remarkable beneficial effects and advantages are achieved.

Description

Technical Field

[0001] The present invention is mainly used in the fields of corporate financing, supply chain industry chain management, risk control, and supervision by regulatory authorities during and after the event, industrial chain layout, credit supervision, resource allocation, etc. It involves technical fields such as big data and artificial intelligence. Specifically, it uses large model technologies such as Transformer and BERT to process and analyze multi-dimensional data to form a large credit chain model, so as to accurately identify and dynamically monitor the risk status of enterprises on the chain. Background Art

[0002] With the rapid development of cutting-edge technologies such as big data, artificial intelligence, and blockchain, as well as the increasing complexity of the global supply chain industry chain and the need for risk management, the identification of risks for enterprises on the chain has ushered in new opportunities and challenges. At present, the existing enterprise risk assessment methods on the market mainly rely on traditional statistical models or simple machine learning models. However, these traditional methods have the following significant defects: Single data dimension: Traditional risk assessment models often rely only on financial data or credit scores, and lack the ability to comprehensively analyze multi-dimensional data. For example, important information such as market transaction data, policy risk data, and corporate evaluation data are often ignored, resulting in incomplete and inaccurate risk assessment results.

[0003] Poor model generalization ability: Existing models are usually general and lack optimization for specific industries, supply chains, and industrial chain types. Different types of supply chains or industrial chains have different risk characteristics, and existing models are difficult to capture these differences, resulting in a decrease in the accuracy of risk assessment.

[0004] Lack of flexibility and scalability: Traditional risk assessment models are usually trained once and are difficult to quickly adjust and optimize based on new data or business needs.

[0005] Insufficient early risk warning and prevention: Traditional methods often focus on post-event processing and lack an effective early warning mechanism, resulting in companies’ slow response when facing risks. Summary of the invention

[0006] A brief summary of one or more aspects is given below to provide a basic understanding of these aspects. This summary is not an exhaustive overview of all conceived aspects, and is neither intended to identify the key or critical elements of all aspects nor to define the scope of any or all aspects. Its only purpose is to give some concepts of one or more aspects in a simplified form as a prelude to a more detailed description that will be given later.

[0007] The purpose of the present invention is to solve the above-mentioned problems and provide a method and system for identifying risks of enterprises on the chain based on the credit chain big model, which provides a more comprehensive, accurate and real-time solution in enterprise risk identification and management, and has significant beneficial effects and advantages.

[0008] The technical solution of the present invention is: the present invention discloses a method for identifying risks of enterprises on the chain based on a credit chain model, the method comprising: Step 1: Data preprocessing: preprocess the input data; Step 2: Matching model: Match the appropriate credit chain model according to the supply chain or industrial chain type for subsequent risk prediction; Step 3: Load weight parameters: load the trained weights and parameters of the corresponding credit chain model; Step 4: Input data calculation and reasoning: import the pre-processed input data into the credit chain model with loaded parameters; Step 5: Output risk prediction results: Based on the calculation results of the credit chain model, output the risk value vector of the enterprises on the chain.

[0009] According to an embodiment of the method for identifying risks of enterprises on the credit chain based on the credit chain model of the present invention, after step 5, the following steps are further included: Step 6: Receive new data: Get the newly arrived data from the set data stream source to prepare for the update of the credit chain model; Step 7: Update parameters through online learning: Use new data combined with online learning algorithms to update the parameters of the credit chain model in real time.

[0010] According to an embodiment of the method for identifying risks of enterprises on the credit chain based on the credit chain model of the present invention, the training process of the credit chain model includes: Step 1: Data processing: Clean and standardize the collected raw data and convert it into a format suitable for model training; Step 2: Pre-training: Convert the processed text into vector representation to provide initial features for model training; Step 3: General model training: Based on the pre-training results, a basic model suitable for on-chain enterprise risk assessment is trained; Step 4: Model fine-tuning and special model training: Based on the general model, train a special model based on the chain characteristics.

[0011] According to an embodiment of the method for identifying risks of enterprises on the credit chain based on the credit chain model of the present invention, step 4 further includes: For on-chain enterprises that need to undergo risk assessment, the collected and cleaned data will be input into the corresponding model with loaded weight parameters. The model uses the previously constructed Transformer encoder layer, pooling layer, attention pooling layer, and convolution layer structure to perform layer-by-layer feature extraction and fusion operations and reasoning on the input data to mine the risk-related feature information contained in the data.

[0012] According to an embodiment of the method for identifying risks of enterprises on the credit chain based on the credit chain model of the present invention, step 5 further includes: After the model's inference operation, the result is processed using the scaled and offset Tanh function in the output layer, and the predicted value, i.e., the risk value, is output. The risk value is presented in the form of a vector with a value range of 0-9. Finally, a multidimensional vector of risk values ​​corresponding to multiple enterprises in the input chain is output.

[0013] According to an embodiment of the method for identifying risks of enterprises on the credit chain based on the credit chain model of the present invention, step 7 further includes: The Adam algorithm of online learning is used to update the model parameters.

[0014] The present invention also discloses a system for identifying risks of enterprises on the credit chain based on a large credit chain model, the system comprising: A data preprocessing module, used for preprocessing input data; Model matching module, used to match the appropriate credit chain model according to the supply chain or industrial chain type, so as to be used for subsequent risk prediction; Parameter loading module, used to load the weights and parameters corresponding to the credit chain model; The data import module is used to import the input data into the credit chain model with loaded parameters; The prediction output module is used to output the risk value vector of the enterprises on the chain according to the calculation results of the credit chain big model.

[0015] The present invention also discloses a computer system for identifying on-chain enterprise risks based on a credit chain big model, comprising a memory, a processor, and program instructions stored in the memory for execution by the processor, wherein the processor executes the program instructions to implement the steps of the on-chain enterprise risk identification method based on the credit chain big model as described above.

[0016] The present invention also discloses a computer-readable storage medium for on-chain enterprise risk identification based on the credit chain big model, which stores program instructions executable by a processor to implement the steps of the on-chain enterprise risk identification method based on the credit chain big model as described above.

[0017] The present invention also discloses a computer program product, including a computer program, which, when executed by a processor, implements the steps of the on-chain enterprise risk identification method based on the credit chain big model as described above.

[0018] Compared with the prior art, the present invention has the following beneficial effects: the present invention uses a data set of multi-dimensional information to retrain the large model to form a general credit vertical model specifically for the risk assessment of on-chain enterprises. The model not only has powerful data processing capabilities and can capture long-distance dependencies between data, but also has high flexibility and scalability, and can perform personalized risk assessments for different types of on-chain enterprises.

[0019] Its beneficial effects are mainly reflected in the following aspects: Multi-dimensional data fusion: The present invention achieves a comprehensive assessment of enterprise risks by integrating multi-dimensional information. This multi-dimensional data fusion overcomes the defect of single data dimension in the prior art and provides a more comprehensive and accurate risk assessment result.

[0020] Dynamic adaptability: This paper adopts a deep learning model based on the stacking of Transformer encoder layers, which can learn and update the latest data in real time, and promptly reflect market dynamics and policy changes. This dynamic adaptability improves the real-time and accuracy of the model in identifying corporate risks, allowing risk assessment to quickly respond to changes in the external environment.

[0021] Flexibility and scalability: The credit chain model of the present invention has high flexibility and scalability, and can be fine-tuned according to the needs, data, and business processes of specific scenarios. This flexibility allows the model to conduct customized risk assessments for different types of chain enterprises, providing a personalized risk assessment solution.

[0022] Capturing long-distance dependencies: Through the self-attention mechanism, the present invention can capture long-distance dependencies between data and better understand the complex relationships between enterprises in the supply chain or industrial chain. This capability improves the accuracy of risk assessment and enables the model to more deeply analyze and predict the risk status of enterprises.

[0023] Real-time risk warning: The model of the present invention can monitor the credit risk and operation status of enterprises in real time, and promptly discover violations and potential risks. This real-time monitoring and early warning mechanism improves the timeliness and effectiveness of supervision, and helps enterprises and regulatory authorities take effective supervision measures during the event to prevent the spread of risks.

[0024] Intelligence and efficiency: This invention uses artificial intelligence and big model technology to achieve accurate assessment of risks of enterprises on the chain, mine multi-dimensional data, improve risk prevention and control and regulatory effectiveness, and has high flexibility and data processing capabilities.

[0025] Cross-industry and cross-regional risk management: The credit chain model constructed by this invention can integrate and analyze data from different industries and regions to provide a unified risk assessment standard. This cross-industry and cross-regional comprehensive analysis enables enterprises to fully understand the risk status of different links and regions in the supply chain or industrial chain and formulate a more comprehensive risk management strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The above features and advantages of the present invention can be better understood after reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings. In the drawings, the components are not necessarily drawn to scale, and components with similar related properties or features may have the same or similar reference numerals.

[0027] Figure 1 A flow chart of the model training process in one embodiment of the method for identifying risks of enterprises on the credit chain based on the credit chain big model of the present invention is shown.

[0028] Figure 2 A flow chart of an embodiment of a method for identifying risks of enterprises on a chain based on a credit chain macro model of the present invention is shown.

[0029] Figure 3 A schematic diagram of an embodiment of a system for identifying risks of enterprises on a chain based on a credit chain macro model of the present invention is shown. DETAILED DESCRIPTION

[0030] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. Note that the aspects described below in conjunction with the accompanying drawings and specific embodiments are only exemplary and should not be construed as limiting the scope of protection of the present invention in any way.

[0031] The training process of the credit chain model is as follows: Figure 1 As shown in the figure, based on the big model, the model is retrained using a data set of multi-dimensional information to adjust the model parameters, optimize the big model into a general credit vertical model that is more suitable for the specific needs of risk assessment and early warning of on-chain enterprises, and realize comprehensive risk identification of on-chain enterprise collections. The specific steps of the training process are detailed below.

[0032] Step 1: Data processing: Clean and standardize the collected raw data and convert it into a format suitable for model training.

[0033] First, we collect multi-dimensional information data sets from multiple channels, including social credit data, market transaction data, policy risk data, and enterprise evaluation data. Next, we clean the text data to remove HTML tags, special characters, and stop words. We use the HanLP tool to segment Chinese text. Then, we standardize the text, unify the format, and eliminate the difference between full-width and half-width characters to ensure data consistency and the effectiveness of model training. Enterprises will be classified according to the type of supply chain or industrial chain to which they belong, and the risk value of the enterprise will be annotated using annotation tools.

[0034] The collected data includes structured data and unstructured data. The general credit model mainly includes the following data sets: public credit information, market credit information, and financial credit information data.

[0035] The data set introduced by the dominant chain credit model also includes: core enterprise financial data, supply chain industry chain management capabilities, industry status and influence, transaction data between core enterprises and upstream and downstream enterprises, and industry-specific risk factors.

[0036] The decentralized chain credit model introduces data sets including: technology development trends, cooperation history and stability, innovation and R&D capabilities.

[0037] The data set introduced by the platform-based chain credit model also includes: platform operation data, policy and compliance data.

[0038] Step 2: Pre-training: Convert the processed text into vector representation to provide initial features for model training.

[0039] First, the cleaned text is encoded using the BERT-Base-Chinese model to obtain context-related vector representations of words and sentences. Among them, word embedding obtains the embedding vector of each word in the text through the BERT-Base-Chinese model, that is, mapping the words into a multi-dimensional vector space so that words with similar semantics are closer in the vector space. Sentence embedding obtains the embedding vector of each sentence in the text through the Sentence-BERT model to reflect the overall semantic characteristics of the sentence.

[0040] Finally, the sentence embedding generated by Sentence-BERT and the word embedding generated by BERT-Base-Chinese are weightedly summed, and the final feature vector is combined according to certain weight rules. This feature vector integrates semantic information at the word and sentence levels for subsequent model training.

[0041] Step 3: General model training: Based on the pre-training results, a basic model suitable for on-chain enterprise risk assessment is trained.

[0042] First, use the pre-trained data to build a model architecture based on the Transformer encoder layer, stacking multiple Transformer encoder layers, each of which contains a self-attention mechanism and a feedforward neural network. The self-attention mechanism allows the model to pay attention to other positions in the input sequence when processing data at a certain position. For example, when analyzing a certain transaction data of an enterprise, it can be associated with other related transactions or other aspects of the enterprise data in the same period, thereby capturing long-distance dependencies between data. Within each Transformer encoder layer, residual connections (adding the input and output of this layer helps gradient propagation and avoids the gradient vanishing problem) and layer normalization (normalizing the input of each layer to make the data distribution more stable and facilitate model training) are used to stabilize the training process and help the model learn deeper features.

[0043] Then, after the Transformer encoder, add a pooling layer (aggregate features to reduce data dimensions), an attention pooling layer (further filter and focus on important features), and a convolutional layer (extract local features) to further extract or fuse features, so that the model can mine useful information from multiple angles for risk assessment. In the early stage of training, the Warmup strategy is used to stabilize the model, and then the Cosine Annealing strategy is used to adjust the learning rate periodically. L2 regularization is applied to the model to avoid excessive reliance on certain features, and Dropout or DropConnect is applied to some layers to improve the generalization ability of the model. Distributed technologies such as data parallelism and model parallelism are used to accelerate the training process, and mixed precision training technology is used to reduce memory usage and accelerate calculation speed. In the output layer, the scaled and offset Tanh function is used to output the predicted value, that is, the risk value. The value range of the vector is 0-9, and finally the n-dimensional vector of the risk value corresponding to the input chain enterprise is output, so as to complete the construction of the general model's preliminary assessment capability of the chain enterprise risk.

[0044] Step 4: Model fine-tuning and special model training: Based on the general model, train a special model based on the chain characteristics.

[0045] Initialize the dedicated model with the weights of the general model, and use the parameters trained by the general model as the initial parameters of the dedicated model. This can speed up the training convergence and utilize some of the general features learned by the general model. Based on the general model, the number, type or parameters of the network layer can be adjusted according to the needs of the dedicated model. Increase or decrease the convolution layer (if you feel the need to strengthen local feature extraction, increase it, otherwise reduce it), pooling layer (adjust the degree of feature aggregation), fully connected layer (change the way of feature fusion and output), etc. During the fine-tuning process, adjust the learning rate to adapt to the characteristics of the new data set. At the same time, according to the characteristics of the chain type, adjust or enhance the feature extraction layer in the model, including adding industry-specific features or adjusting the weights of existing features. Use different training data sets and parameters to implement three chain-specific models, namely the dominant chain credit model, the decentralized chain credit model, and the platform chain credit model. For example, the dominant chain credit model is applied to scenarios where the core enterprise is in an advantageous position in the supply chain or industrial chain. The core enterprise as a whole controls key information such as prices, orders, and goods of upstream and downstream enterprises, and has an absolute dominant position in the supply chain or industrial chain. Upstream and downstream enterprises are highly concentrated, and the focus will be on the financial data and supply chain and industrial chain management capabilities of the core enterprise; the decentralized chain credit model is applied to scenarios where there is no absolute dominant enterprise in the supply chain or industrial chain. The concentration of upstream and downstream enterprises is not high, and attention will be paid to technological development trends and cooperation history; the platform chain credit model is applied to platform-type enterprises or units. The platform brings together various upstream and downstream information, including e-commerce, logistics, government procurement and other platforms.

[0046] Figure 2 The flow chart of an embodiment of the method for identifying risks of enterprises on the chain based on the credit chain model of the present invention is shown. Figure 2 As shown, the implementation steps of the method of this embodiment are described in detail as follows.

[0047] Step 1: Data preprocessing: Preprocess the prepared input data, such as data cleaning, normalization, etc.

[0048] Perform text cleaning to remove HTML tags, special characters, stop words, etc. Use HanLP tools to perform Chinese text segmentation. Then perform text standardization to convert all texts into a unified format to eliminate the differences between full-width characters and half-width characters.

[0049] Step 2: Matching model: Match the appropriate model according to the supply chain or industrial chain type for subsequent risk prediction.

[0050] Determine the type of supply chain or industrial chain. For conventional supply chains or industrial chains that do not show obvious specific chain characteristics, select a trained general credit model. For supply chains or industrial chains dominated by core enterprises, select a dominant chain credit model; for decentralized supply chains or industrial chains, select a decentralized chain credit model; for platform supply chains or industrial chains, select a platform chain credit model.

[0051] Step 3: Load weight parameters: Load the trained weights and parameters of the corresponding model to prepare for subsequent reasoning.

[0052] From the file saved after model training, accurately read and load the previously trained corresponding model so that the model is in a state that can be used for inference and prediction.

[0053] Step 4: Input data for computational reasoning: Input the prepared data into the model with loaded parameters so that the model can perform computational reasoning.

[0054] For the on-chain enterprises that need to conduct risk assessment, the collected and cleaned data will be input into the corresponding model (general model or special model) with loaded weight parameters. The model will use the previously constructed Transformer encoder layer (each layer uses the self-attention mechanism to focus on the long-distance dependencies between data, and uses residual connections and layer normalization to stabilize the training and learning features, as well as feedforward neural network to process data), pooling layer, attention pooling layer, convolution layer and other structures to perform layer-by-layer feature extraction, fusion and other operational reasoning on the input data to mine the risk-related feature information contained in the data.

[0055] Step 5: Output risk prediction results: Output the on-chain enterprise risk value vector based on the model calculation results.

[0056] After the model's reasoning operation, the result is processed using the scaled and offset Tanh function in the output layer to output the predicted value, which is the risk value. This risk value is presented in the form of a vector, and its value range is set to 0-9. The final output is an n-dimensional vector of the risk values ​​corresponding to the n companies in the input chain. For example, if the data of 10 companies in a supply chain or industrial chain is input, after the model is processed, a 10-dimensional vector will be output. The value of each dimension corresponds to the risk value of a company. The closer the value is to 9, the higher the risk, and the closer it is to 0, the lower the risk. This intuitively displays the risk status of each company in the chain for subsequent analysis and decision-making.

[0057] On the basis of the above embodiment, the following steps 6 and 7 may be added for further optimization.

[0058] Step 6: Receive new data: Get the newly arrived data from the set data stream source to prepare for model update.

[0059] Through the interface and data transmission channel corresponding to the set data stream source, the newly arrived data is received in real time, and the received data is preliminarily processed such as format verification and integrity check to ensure the accuracy and availability of the data.

[0060] Step 7: Update parameters through online learning: Use new data combined with online learning algorithms to update model parameters in real time to improve the adaptability of the model.

[0061] The online learning algorithm Adam (Adam Adaptive Moment Estimation, an optimization algorithm for training neural networks) is used to update the model parameters. The Adam algorithm can adaptively adjust the update step size of the model parameters according to the newly arrived data points, and can more effectively optimize the model parameters when facing data of different scales and characteristics. In specific applications, when new data is received, the gradient change of the new data on each model parameter is calculated based on the Adam algorithm, and then the model parameters are updated according to certain rules. At the same time, an exponential decay strategy is added to adjust the learning rate, that is, as new data is continuously input, the learning rate is gradually reduced, so that the model can balance the impact of old and new data on the model in the process of learning new data, avoid over-reliance on new data and ignore the previously learned knowledge structure, and ensure that the model can not only adapt to new risk changes in a timely manner, but also maintain the effective use of previous data features, thereby continuously improving the accuracy and adaptability of the model to the risk assessment of enterprises on the chain.

[0062] Through the above complete process, different models can be used to predict the risks of enterprises on the chain, and the models can be continuously optimized through dynamic learning to make them better adapt to changes in the risk status of enterprises and meet the needs of risk identification and management of enterprises on the chain.

[0063] Figure 3 The principle of an embodiment of the system for identifying risks of enterprises on the chain based on the credit chain model of the present invention is shown. Figure 3 The system of this embodiment includes: a data preprocessing module, a model matching module, a parameter loading module, a data import module, and a prediction output module.

[0064] The data preprocessing module is used to preprocess the input data. Figure 2 Step 1 of the method embodiment shown is not repeated here.

[0065] The model matching module is used to match the appropriate credit chain model according to the supply chain or industrial chain type for subsequent risk prediction. Figure 2 Step 2 of the method embodiment shown is not repeated here.

[0066] The parameter loading module is used to load the weights and parameters corresponding to the credit chain model. Figure 2 Step 3 of the method embodiment shown is not repeated here.

[0067] The data import module is used to import the input data into the credit chain model with loaded parameters. Figure 2 Step 4 of the method embodiment shown is not repeated here.

[0068] The prediction output module is used to output the risk value vector of the enterprise on the chain according to the calculation results of the credit chain model. Figure 2 Step 5 of the method embodiment shown is not repeated here.

[0069] In addition, the present invention also discloses a computer system for on-chain enterprise risk identification based on the credit chain big model, including a memory, a processor and program instructions stored in the memory for the processor to run, wherein the processor executes the program instructions to implement the steps of the on-chain enterprise risk identification method based on the credit chain big model as described above.

[0070] In addition, the present invention also discloses a computer-readable storage medium for on-chain enterprise risk identification based on the credit chain big model, which stores program instructions executable by a processor to implement the steps of the on-chain enterprise risk identification method embodiment based on the credit chain big model as described above.

[0071] In addition, the present invention also discloses a computer program product, including a computer program, which, when executed by a processor, implements the steps of the embodiment of the on-chain enterprise risk identification method based on the credit chain big model as described above.

[0072] Overall, the present invention has the following advantages over traditional technologies: the credit chain model can reflect market dynamics and policy changes in real time by constantly updating and learning the latest data. For example, when the market environment fluctuates or new policies are introduced, the model will automatically adjust its parameters and analysis logic to maintain the accuracy and real-time nature of enterprise risk identification. This dynamic adjustment capability enables the credit chain model to not only accurately identify current risks, but also to warn of potential risks in the future.

[0073] The intelligent features of the model are reflected in its comprehensive analysis of multiple data sources and its autonomous learning capabilities. By integrating social credit data and market transaction data, the model can comprehensively assess the credit status of enterprises; combined with policy risk data, the model can predict the potential impact of policy changes on enterprises; using enterprise evaluation data, the model can further refine the analysis of individual enterprises and provide more targeted risk identification results.

[0074] (I) Application and scenarios The main uses and application scenarios of the present invention can be divided into two aspects: one is for chain leader enterprises, and the other is for regulatory authorities.

[0075] For chain leader enterprises, the present invention has important application value in dynamic assessment of credit risk, supply chain and industrial chain management, and risk control strategies.

[0076] 1. Dynamically assess credit risk: In the process of financing, the chain leader enterprise needs to show the credit status and risk level of itself and the upstream and downstream enterprises in the supply chain and industrial chain to the financial institutions. By using the credit chain big model of the present invention, the chain leader enterprise can dynamically and accurately assess the credit risk of itself and its supply chain or industrial chain partners, and provide a detailed risk identification report.

[0077] 2. Supply chain and industrial chain management: In the management of supply chain and industrial chain, chain leader enterprises need to pay close attention to the operating status and risk level of enterprises in each link of the supply chain and industrial chain. Through the intelligent data annotation system and the credit chain big model, the present invention can monitor the risk status of each enterprise in the supply chain and industrial chain in real time, discover potential risks in time and take preventive measures. For example, when the credit status of a supplier deteriorates, the chain leader enterprise can quickly adjust the procurement strategy and find alternative suppliers to avoid the supply chain or industrial chain rupture caused by problems with a single supplier.

[0078] 3. Risk control: Risk control is a key link in the management of chain owner enterprises. Using the present invention, chain owner enterprises can comprehensively analyze various risks in the supply chain and industrial chain, including financial risks, operational risks, market risks, etc. Through in-depth mining and analysis of risk data, chain owner enterprises can formulate scientific risk control strategies to ensure the stability and sustainability of enterprise operations. In addition, the present invention can also help chain owner enterprises establish a complete risk early warning mechanism, discover and respond to potential risks in advance, and enhance the overall risk resistance of the enterprise.

[0079] For regulatory authorities, the present invention has important application value in in-process and post-process supervision, supply chain and industrial chain layout optimization, credit supervision and resource allocation.

[0080] 1. Supervision during and after the event: Regulatory authorities need to continuously supervise enterprises in the market to ensure their compliance operations. Through the present invention, regulatory authorities can monitor the credit risk and operating status of enterprises in real time, and promptly discover violations and potential risks. The accurate risk identification report provided by the present invention can help regulatory authorities take effective regulatory measures during the event to prevent the spread of risks; conduct in-depth investigations and handling afterwards to improve the timeliness and effectiveness of supervision.

[0081] 2. Optimization of supply chain and industrial chain layout: When formulating industrial policies and layout plans, regulatory authorities need to understand the credit status and risk level of enterprises in each link of the supply chain and industrial chain. Using this invention, regulatory authorities can comprehensively analyze the operating data of each enterprise in the supply chain and industrial chain, and identify the key nodes and vulnerable links in the supply chain and industrial chain. Through the overall risk assessment of the supply chain and industrial chain, regulatory authorities can formulate scientific industrial policies, optimize the layout of the supply chain and industrial chain, and promote the healthy development of the supply chain and industrial chain.

[0082] 3. Credit supervision: The regulatory authorities are responsible for formulating and implementing corporate credit supervision policies. Through the intelligent data annotation system and the credit chain model, the present invention can accurately assess the credit risk of enterprises and provide detailed credit reports. Based on these reports, regulatory authorities can conduct credit rating and classification management of enterprises, formulate corresponding credit supervision measures, and promote the optimization of the market credit environment.

[0083] 4. Resource allocation: In the process of resource allocation, the regulatory authorities need to consider the credit status and risk level of the enterprise. Through the risk identification report provided by the present invention, the regulatory authorities can scientifically evaluate the resource needs and risk tolerance of the enterprise and optimize the resource allocation plan. This can not only improve the efficiency of resource allocation, but also effectively prevent resource waste and risk accumulation, and promote the rational use of resources and sustainable economic development.

[0084] (II) Problem Solving 1. Dynamics of on-chain enterprise risks The risks of enterprises on the chain are dynamic and are affected by various factors such as the market environment, economic policies and the business conditions of the enterprises themselves. Traditional static risk assessment methods cannot reflect these changes in a timely manner, which easily leads to a lag in risk control. This invention uses advanced technologies such as artificial intelligence and large models to construct a large credit chain model that can dynamically monitor and analyze enterprise risks. Through real-time updates and adaptive adjustments, the model can quickly respond to changes in the external environment and internal operating conditions, provide accurate risk warnings and assessments, and ensure that enterprises can take timely countermeasures.

[0085] 2. Traditional methods are inefficient Traditional risk identification and assessment methods rely on manual analysis and manual operation, which are inefficient and prone to errors, and cannot meet the real-time and accuracy requirements of modern enterprises and regulatory authorities. This invention uses artificial intelligence and large model technology to automatically process data and identify risks, greatly improving efficiency and accuracy. This makes the risk identification process more intelligent and automated, reduces human intervention, and improves the efficiency of risk management.

[0086] 3. Challenges of cross-industry and cross-regional risk management In the modern economic environment, the business, supply chain and industrial chain of enterprises often span multiple industries and regions, which increases the complexity of risk management. The market environment, laws and regulations, and economic conditions in different regions are different, and traditional risk management methods are difficult to fully cover and effectively respond to. The present invention can integrate and analyze data from different industries and regions through the constructed credit chain model to provide a unified risk assessment standard. Through this cross-industry and cross-regional comprehensive analysis, enterprises can fully understand the risk status of different links and regions in the supply chain and industrial chain, and formulate a more comprehensive risk management strategy.

[0087] 4. Insufficient early risk warning and prevention Traditional risk management methods often focus on post-processing and lack an effective early warning mechanism, which leads to slow response of enterprises in the face of risks and difficulty in taking preventive measures in a timely manner. The present invention uses a credit chain model constructed by artificial intelligence and large model technology to achieve real-time monitoring and early warning of risks in enterprises, supply chains and industrial chains. Through deep learning and analysis of a large amount of historical data and real-time data, the model can identify potential risk signals and trends, and warn of possible risk events in advance. This early warning mechanism helps enterprises take preventive measures before risks occur, reducing the impact and losses of risks.

[0088] Advantages or benefits compared to existing technologies Compared with the prior art, the present invention has significant beneficial effects and advantages, which are specifically reflected in the following aspects: 1. Dynamic and real-time Existing risk identification methods are usually static and cannot reflect changes in the market environment, economic policies and corporate credit status in a timely manner. The present invention uses advanced technologies such as artificial intelligence and large models to build a dynamically evolving credit chain large model. By real-time monitoring and analysis of multi-dimensional data, the model can be updated and adjusted in a timely manner to accurately reflect the risk status of the enterprise. This dynamism and real-time nature ensures the timeliness of risk identification and helps enterprises and regulators to quickly respond to changing risk environments.

[0089] 2. Comprehensiveness and precision Traditional risk identification methods often rely on a single data source or simple analysis methods, resulting in incomplete and inaccurate risk assessment results. The model can deeply mine and analyze these data to provide accurate risk identification results. This comprehensiveness and accuracy greatly improves the reliability and effectiveness of risk assessment.

[0090] 3. Intelligence and efficiency Existing technologies usually require a lot of manual intervention in the risk identification process, which is inefficient and error-prone. The present invention uses artificial intelligence and large model technology to achieve accurate assessment of risks of enterprises on the chain, mine multi-dimensional data, improve risk prevention and control and regulatory efficiency, and has high flexibility and data processing capabilities.

[0091] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0092] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0093] The various illustrative logic blocks, modules, and circuits described in conjunction with the embodiments disclosed herein may be implemented or executed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in an alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0094] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read and write information from / to the storage medium. In an alternative, a storage medium may be integrated into a processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components.

[0095] In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented as a computer program product in software, each function may be stored on or transmitted by a computer-readable medium as one or more instructions or codes. Computer-readable media include both computer storage media and communication media, including any media that facilitates the transfer of a computer program from one place to another. Storage media may be any available media that can be accessed by a computer. As an example and not limitation, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, disk storage or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer. Any connection is also properly referred to as a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwaves, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwaves are included in the definition of the medium. Disk and disc as used herein include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc, where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0096] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but should be granted the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying risks of enterprises on the credit chain based on a large credit chain model, characterized in that: Methods include: Step 1: Data preprocessing: preprocess the input data; Step 2: Matching model: Match the appropriate credit chain model according to the supply chain or industrial chain type for subsequent risk prediction; Step 3: Load weight parameters: load the trained weights and parameters of the corresponding credit chain model; Step 4: Input data calculation and reasoning: import the pre-processed input data into the credit chain model with loaded parameters; Step 5: Output risk prediction results: Based on the calculation results of the credit chain model, output the risk value vector of the enterprises on the chain.

2. The method for identifying risks of enterprises on the credit chain based on the credit chain big model according to claim 1 is characterized in that: After step 5, also include: Step 6: Receive new data: Get the newly arrived data from the set data stream source to prepare for the update of the credit chain model; Step 7: Update parameters through online learning: Use new data combined with online learning algorithms to update the parameters of the credit chain model in real time.

3. The method for identifying risks of enterprises on the credit chain based on the credit chain big model according to claim 1 is characterized in that: The training process of the credit chain model includes: Step 1: Data processing: Clean and standardize the collected raw data and convert it into a format suitable for model training; Step 2: Pre-training: Convert the processed text into vector representation to provide initial features for model training; Step 3: General model training: Based on the pre-training results, a basic model suitable for on-chain enterprise risk assessment is trained; Step 4: Model fine-tuning and special model training: Based on the general model, train a special model based on the chain characteristics.

4. The method for identifying risks of enterprises on the credit chain based on the credit chain big model according to claim 1 is characterized in that: Step 4 further includes: For on-chain enterprises that need to undergo risk assessment, the collected and cleaned data will be input into the corresponding model with loaded weight parameters. The model uses the previously constructed Transformer encoder layer, pooling layer, attention pooling layer, and convolution layer structure to perform layer-by-layer feature extraction and fusion operations and reasoning on the input data to mine the risk-related feature information contained in the data.

5. The method for identifying risks of enterprises on the credit chain based on the credit chain big model according to claim 1 is characterized in that: Step 5 further includes: After the model's inference operation, the result is processed using the scaled and offset Tanh function in the output layer, and the predicted value, i.e., the risk value, is output. The risk value is presented in the form of a vector with a value range of 0-9. Finally, a multidimensional vector of risk values ​​corresponding to multiple enterprises in the input chain is output.

6. The method for identifying risks of enterprises on the credit chain based on the credit chain big model according to claim 2 is characterized in that: Step 7 further includes: The Adam algorithm of online learning is used to update the model parameters.

7. A system for identifying risks of enterprises on the credit chain based on a large credit chain model, characterized in that: The system includes: A data preprocessing module, used for preprocessing input data; Model matching module, used to match the appropriate credit chain model according to the supply chain or industrial chain type, so as to be used for subsequent risk prediction; Parameter loading module, used to load the weights and parameters corresponding to the credit chain model; The data import module is used to import the input data into the credit chain model with loaded parameters; The prediction output module is used to output the risk value vector of the enterprises on the chain according to the calculation results of the credit chain big model.

8. A computer system for identifying risks of enterprises on the credit chain based on a large credit chain model, characterized in that: It comprises a memory, a processor and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement the steps of the on-chain enterprise risk identification method based on the credit chain big model as described in any one of claims 1 to 6.

9. A computer-readable storage medium for identifying on-chain enterprise risks based on a credit chain model, characterized in that: It stores program instructions executable by a processor to implement the steps of the on-chain enterprise risk identification method based on the credit chain model as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the on-chain enterprise risk identification method based on the credit chain big model as described in any one of claims 1 to 6 are implemented.

Images