A green loan risk assessment and pricing system

Abstract

The application discloses a kind of green loan risk assessment and pricing system, comprising: data collection module collects the financial data of green loan application enterprise, industry chain information, environmental data, and the historical data of green loan and bank green financial product, and carbon emission and carbon absorption related data of loan enterprise;Data processing module carries out data preprocessing;Feature extraction module extracts features to data and constructs feature vector;Model training module trains the deep learning model of risk assessment and pricing through feature vector;Evaluation pricing module uses model to assess the risk of new loan application, carbon absorption potential and carbon balance risk, and combines environmental benefits and key technical factors of green capacity to price loans, according to the evaluation results and loan pricing suggestions output by the model, develop green loan business strategy.The application can comprehensively, carefully, accurately and scientifically assess the risk of green loan.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of financial technology, particularly the field of green finance, and specifically relates to a green loan risk assessment and pricing system. BACKGROUND

[0002] With the global concern about climate change, green finance has become an important development direction of the financial industry. As an important part of green finance, the scientificity and accuracy of the risk assessment and pricing mechanism of green loans are crucial to the sustainable development of banks.

[0003] However, the existing green loan risk assessment and pricing system often relies on traditional credit assessment models, lacks comprehensive consideration of environmental factors and policy guidance, and cannot meet the rapid development needs of green loan business.

[0004] At the same time, in the current risk rating system of the banking industry, the traditional five-level risk loan classification mainly focuses on single loan business, and its evaluation range is relatively narrow. This single perspective of risk assessment method cannot fully reflect the complexity and potential risks of green financial business. In addition, the existing green loans of the banking industry only focus on the evaluation of carbon emissions in one field, and lack of attention to the carbon absorption field. SUMMARY

[0005] In view of the shortcomings of the existing banking industry that only uses historical financial income and expenditure data of the applicant enterprise to assess loan credit, and only focuses on single loan business while lacking attention to the carbon absorption field, the present application proposes a green loan risk assessment and pricing system. By collecting the financial data, industry chain information, and environmental data of green loan applicant enterprises, collecting the historical data of green loans and bank green financial products, and collecting the carbon emissions and carbon absorption of loan enterprises, the system can realize comprehensive risk rating of green financial product portfolio based on multiple dimensions of data, provide more comprehensive and detailed risk assessment results for banks, and expand the evaluation range to the entire green financial product chain to build a more comprehensive and accurate risk assessment framework. At the same time, the system adds the carbon absorption field to the evaluation model, assesses the additional carbon absorption of green loan enterprises in detail, and then actively matches the regional carbon absorption enterprises through a deep learning model, so as to realize more scientific artificial intelligence big model green loan risk assessment and pricing.

[0006] The green loan risk assessment and pricing system according to the present application comprises:

[0007] A data collection module for collecting the financial data, industry chain information, and environmental data of green loan applicant enterprises, collecting the historical data of green loans and bank green financial products, and collecting the carbon emissions and carbon absorption related data of loan enterprises;

[0008] Data processing module: for data preprocessing of collected data, the data processing includes: cleaning, denoising, removing outliers and missing values, standardization processing, normalization processing;

[0009] Feature extraction module: for extracting features from data preprocessed data and constructing feature vectors for input of subsequent deep learning model;

[0010] Model training module: for designing a deep learning model for risk assessment and pricing including convolutional layer and recurrent layer, and training the model through the feature vector;

[0011] Evaluation and pricing module: for using the trained model to assess the risk of new loan applications, and carbon absorption potential and carbon balance risk, and combining environmental benefits and key technical factors of green production capacity to price loans, and according to the evaluation results and loan pricing suggestions output by the model, formulating green loan business strategy.

[0012] Further, including model optimization module: for dynamically updating the model according to market changes and policy guidance, and adjusting the loan credit limit and pricing using optimization algorithm according to the risk assessment results, and collecting the carbon emission and carbon absorption data of the enterprise after the loan, for feedback and iterative optimization of the model.

[0013] Further, the formula used in adjusting the loan credit limit and pricing using optimization algorithm according to the risk assessment results is shown in formula (1) as follows:

[0014] (1)

[0015] In the formula, is the loan interest rate, is the base interest rate, is a natural constant, is the risk score, is the market carbon credit price, and is the adjustment parameter.

[0016] Further, the evaluation and pricing module specifically includes: loan pricing according to the comprehensive risk rating, market interest rate and green project yield rate output by the model, wherein the formula used is shown in formula (2) as follows:

[0017] (2)

[0018] In the formula, is the loan pricing, is the comprehensive risk rating Risk_Rating, is the market interest rate Market_Rate, Green_Project_ROI is a green project return on investment, 、 、 、 is a regression coefficient, is an error term.

[0019] Further, the carbon absorption potential evaluation includes evaluating the increased carbon absorption potential of the enterprise through biological carbon sequestration and CCUS technology, and the carbon balance risk evaluation includes using a deep learning model to evaluate the risk according to the carbon emission potential and the carbon absorption potential of the enterprise.

[0020] Further, the formula used in the carbon absorption potential evaluation is shown in the following formula (3):

[0021] T (3)

[0022] In the formula, is the carbon absorption potential, is the forest coverage rate, and T is the technical carbon sequestration capacity, 、 、 is a model parameter;

[0023] The formula used in the carbon balance risk evaluation is shown in the following formula (4):

[0024] (4)

[0025] In the formula, R is the carbon balance risk score, e is a natural constant, is the carbon emission, is the carbon absorption potential, is a policy impact factor, is a market sensitivity factor, 、 、 、 is a weight parameter.

[0026] Further, it also includes a carbon absorption enterprise matching module, which is used to identify and match carbon absorption enterprises in the bank's private domain information database using an artificial intelligence algorithm, wherein the formula used is shown in the following formula (5):

[0027] (5)

[0028] In the formula, is the minimum value in all matching schemes, i.e. the value corresponding to the optimal matching scheme, is the total number of carbon emission enterprises applying for loans, a carbon emission enterprise applying for the i-th loan, a carbon absorption enterprise in the private domain information base of the bank, a distance function between the i-th carbon emission enterprise applying for the loan and the j-th carbon absorption enterprise, a cost function of the j-th carbon absorption enterprise, a trade-off parameter.

[0029] Further, the model training module comprises: introducing a Dropout layer to reduce overfitting, combining an attention mechanism to construct a model, and using a back propagation algorithm to optimize model parameters.

[0030] Compared with the prior art, the present application has the beneficial effects that:

[0031] 1. The present application proposes a green loan risk assessment and pricing system, which collects financial data, industry chain information and environmental data of green loan application enterprises, collects historical data of green loans and bank green financial products, and collects carbon emissions and carbon absorption of loan enterprises, then performs data preprocessing and feature extraction on these data for training of a deep learning model, finally performs risk assessment, carbon absorption potential assessment and carbon balance risk assessment on new loan applications through the trained model, and combines environmental benefits and key technical factors of green production capacity to price the loan, and formulates a green loan business strategy according to the evaluation results and loan pricing suggestions output by the model,

[0032] 2. The present application can realize comprehensive risk rating of green financial product portfolio based on multiple dimensional data, and provide more comprehensive and detailed risk assessment results for the bank; and can expand the evaluation range to the entire green financial product chain to build a more comprehensive and accurate risk assessment framework.

[0033] 3. The present application adds the carbon absorption field to the evaluation model, and detailedly evaluates the green loan enterprise to add carbon absorption, then actively matches the carbon absorption enterprise in the region through the deep learning model, so that a more scientific artificial intelligence big model green loan risk assessment and pricing can be realized. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 A functional module diagram of a green loan risk assessment and pricing system provided by the present application; DETAILED DESCRIPTION

[0035] In order to make the technical means, creative features, purposes and effects realized by the present application easy to understand, the present application is further described below in combination with specific embodiments.

[0036] As Figure 1As shown, the green loan risk assessment and pricing system provided by the application comprises a data collection module, a data processing module, a feature extraction module, a model training module and an evaluation and pricing module. The data collection module collects various data related to green loans, and then sends the data to the data processing module for data preprocessing. After the data processing module completes data preprocessing, it inputs the data into the feature extraction module. The feature extraction module extracts features from the input data and constructs a feature vector, and then inputs the feature vector into the model training module to train the model. After the model training module trains the model, the evaluation and pricing module uses the trained model to evaluate the risk of new loan applications, as well as the carbon absorption potential and carbon balance risk, and combines the key technical factors of environmental benefits and green production capacity to price the loan. Then, according to the evaluation results and loan pricing suggestions output by the model, a green loan business strategy is developed.

[0037] The data collection module is used to collect the financial data, industry chain information and environmental data of the green loan applicant enterprise, collect the historical data of green loans and bank green financial products, and collect the carbon emission and carbon absorption related data of the loan enterprise.

[0038] Specifically, in this module, the initial data of the green loan applicant enterprise and the entire industry chain is first collected, specifically the financial data, industry chain information and environmental data of the green loan applicant enterprise, which specifically includes operating income, net profit, asset-liability ratio, energy consumption, etc.

[0039] Secondly, the historical data of green loans and bank green financial products is collected, such as the green loan data of the past five years exported from the bank internal system, which specifically includes but is not limited to basic data and information such as historical green loan amount, loan period, interest rate level, repayment record, etc.

[0040] Finally, the carbon emission and carbon absorption related data of the loan enterprise is collected, which specifically includes but is not limited to data information such as energy consumption, production process, forest coverage rate, CCUS facilities, etc.

[0041] The data processing module is used to preprocess the collected data, and the data processing includes cleaning, denoising, removing outliers and missing values, standardization processing and normalization processing.

[0042] Specifically, in this module, the collected data is sequentially cleaned, i.e. removing duplicate, erroneous or severely missing records, denoised, outliers and missing values are removed, and then standardized or normalized to ensure that data of different sources and different dimensions can be compared and analyzed on the same scale. This series of data preprocessing can effectively improve the data quality.

[0043] In the standardization processing, the following formula is used:

[0044]

[0045] In the formula, is the standardized data, is the original data, is the mean value, is the standard deviation.

[0046] In other embodiments, other different standardization processing methods can also be used, that is, the Min-Max standardization method, and the formula used is as follows:

[0047]

[0048] By using different standardization processing methods, the influence of different standardization processing methods on the performance of the model can be tested.

[0049] Feature extraction module: used for extracting features from the preprocessed data and constructing a feature vector for the input of the subsequent deep learning model;

[0050] Specifically, the PCA method can be used to extract features from the preprocessed data, and the PCA (principal components analysis) is a principal component analysis technique, also known as principal component analysis, which aims to use the idea of dimensionality reduction to convert multiple indicators into a few comprehensive indicators, so that feature selection and dimensionality reduction can be performed through PCA. In actual operation, PCA can be used to reduce the dimension to 10 main features.

[0051] In the PCA method, the following formula is used:

[0052]

[0053] In the formula, is the principal component score, is the original data, is the mean vector, is the feature vector, is the eigenvalue.

[0054] Further, when extracting features, key features such as loan term, green project type (such as renewable energy, energy-saving building, etc.), carbon emission reduction percentage, and carbon trading price change trend are extracted.

[0055] Model training module: used for designing a deep learning model for risk assessment and pricing including convolutional layers and recurrent layers, and performing model training through the feature vector;

[0056] Specifically, in this embodiment, the deep learning model includes convolutional layers and recurrent layers, i.e., convolutional neural networks (CNN) or recurrent neural networks (RNN), and the input layer contains selected features and the output layer is a risk score. After data preprocessing and feature extraction on the collected data, the deep learning model is trained, and in this process, the Adam optimizer and cross-entropy loss function can be used.

[0057] During the training process, by introducing relevant environmental data as input features, the model can learn the impact of carbon emissions, carbon emission policy changes, carbon market price fluctuations, and other factors on green loan risk. The specific formula used is as follows:

[0058]

[0059] wherein, is the output, x is the input feature vector, is the weight matrix of the convolution kernel, is the bias term, is the activation function.

[0060] During the training process, by continuously adjusting the model parameters such as the weight matrix and the bias term , the prediction error can be minimized.

[0061] Through multiple iterations of training, the model gradually converges and shows good generalization ability on the validation set. The formula used in iterative training is as follows:

[0062]

[0063] wherein, is the loss function, is the true label, is the probability predicted by the model.

[0064] In an optional implementation, the model training module includes: introducing a Dropout layer to reduce overfitting, combining an attention mechanism to build the model, and using a backpropagation algorithm to optimize the model parameters.

[0065] Specifically, in this embodiment, according to the data characteristics and problem complexity, we selected a recurrent neural network combined with an attention mechanism (RNN-Attention) as the basic architecture of the risk assessment and pricing model. This architecture can effectively capture long-term dependencies in time series data and highlight important information through the attention mechanism.

[0066] In addition, in the training process, in addition to introducing the Dropout layer to reduce overfitting, early stopping method or increasing the regularization term can be used to prevent overfitting.

[0067] The evaluation and pricing module is configured to use the trained model to evaluate the risk of a new loan application, and to evaluate the carbon absorption potential and carbon balance risk, and to price the loan in combination with the key technical factors of environmental benefits and green capacity, and to develop a green loan business strategy based on the evaluation results and loan pricing suggestions output by the model.

[0068] Specifically, after using the trained model to evaluate the risk of a new green loan application, the loan can be priced in combination with the key technical factors of environmental benefits and green capacity, and the expression is as follows:

[0069]

[0070] In the formula, is the loan pricing, is the risk score, is the environmental benefit, is the technical factor.

[0071] Further, in an optional embodiment, the evaluation and pricing module specifically includes: according to the comprehensive risk rating output by the model, the market interest rate, and the green project return rate, the loan is priced, and the formula used is as follows formula (2):

[0072] (2)

[0073] In the formula, is the loan pricing, is the comprehensive risk rating, is the market interest rate, is the green project return rate, , , , is the regression coefficient, is the error term.

[0074] wherein, , , , can be obtained by historical data regression, is a random error term.

[0075] In this embodiment, when a new green loan application is submitted, the system will automatically extract relevant features and input them into the trained deep learning model. The model will output a comprehensive risk rating, taking into account the creditworthiness of the loan applicant and the environmental benefits of the green project. The model will also make pricing predictions and recommendations based on current market interest rates, expected returns on green projects (Green_Project_ROI), and other factors. The bank's credit department can quickly make loan approval decisions based on the model's risk assessment results and pricing recommendations, and develop personalized green loan business strategies.

[0076] Through this embodiment, the present application can accurately assess and price green loan risks, improve the efficiency and security of bank green loan business, and promote the sustainable development of green finance.

[0077] In an optional implementation, the carbon absorption potential assessment includes assessing the increased carbon absorption potential of the enterprise through biological carbon sequestration and CCUS technology, and the carbon balance risk assessment includes using a deep learning model to assess risks based on the carbon emission potential and carbon absorption potential of the enterprise.

[0078] In an optional implementation, the formula used in the carbon absorption potential assessment is as shown in the following formula (3):

[0079] T (3)

[0080] In the formula, is the carbon absorption potential, is the forest coverage rate, and T is the technical carbon sequestration capacity, , , is a model parameter;

[0081] wherein, , , can be pre-set by an artificial intelligence model, for example = 0.5, = 1.2, = 0.3.

[0082] The formula used in the carbon balance risk assessment is as shown in the following formula (4):

[0083] (4)

[0084] In the formula, R is the carbon balance risk score, e is the natural constant, is the carbon emission, is the carbon absorption potential, is a policy influence factor, Market sensitivity factor, , , , Weight parameter.

[0085] Wherein, , , , Can be pre-set by artificial intelligence model, for example = 0.4, = 0.3, = 0.2, = 0.1.

[0086] In an alternative embodiment, it further comprises a carbon absorption enterprise matching module, which is used to identify and match carbon absorption enterprises in the bank's private domain information base by using artificial intelligence algorithm, wherein the formula used is shown in the following formula (5):

[0087] (5)

[0088] Wherein, Is the minimum value in all matching schemes, that is, the value corresponding to the optimal matching scheme, Is the total number of carbon emission enterprises applying for loans, Is the i-th carbon emission enterprise applying for loans, Is the j-th carbon absorption enterprise in the bank's private domain information base, Is the distance function between the i-th carbon emission enterprise applying for loans and the j-th carbon absorption enterprise, Is the cost function of the j-th carbon absorption enterprise, Is the trade-off parameter. Wherein, The value of can be 0.6.

[0089] Specifically, the matching of carbon emission enterprises and carbon absorption enterprises can be realized through the carbon trading market. Carbon trading market is an effective means to reduce carbon emissions through market mechanism. The government sets an emission cap (quota) for enterprises, and enterprises that exceed the quota need to buy additional carbon rights on the carbon market, while enterprises that do not use up the quota can sell the remaining carbon rights, thereby promoting emission reduction.

[0090] Wherein, the operation mechanism of carbon trading market is as follows:

[0091] 1. Quota management: the government sets a greenhouse gas emission cap (quota) for enterprises, and the amount of carbon that enterprises can emit per year is limited.

[0092] 2. Carbon trading: Companies that exceed their quotas need to buy additional carbon credits on the carbon market, while companies that do not use up their quotas can sell the remaining carbon credits.

[0093] 3. Carbon sink projects: Companies can offset part of their carbon emissions by investing in carbon sink projects such as afforestation.

[0094] Through the pairing method in the pairing module designed in this embodiment, the bank can quickly find the optimal pairing scheme in the carbon trading market, thereby promoting the activity of the carbon trading market and promoting the development of green finance.

[0095] In an optional implementation, the system further includes a model optimization module: for dynamically updating the model according to market changes and policy guidance, and adjusting the loan credit limit and pricing using an optimization algorithm according to the risk assessment results, as well as collecting carbon emission and carbon absorption data of the enterprise after the loan, for feedback and iterative optimization of the model.

[0096] where the following iterative formula can be used:

[0097]

[0098] where, is the updated model parameter, is the old model parameter, is the learning rate, is the gradient of the loss function. Where, The value of can be 0.01.

[0099] In an optional implementation, the formula used in adjusting the loan credit limit and pricing using an optimization algorithm according to the risk assessment results is as shown in the following formula (1):

[0100] (1)

[0101] where, is the loan interest rate, is the base interest rate, is a natural constant, is the risk score, is the market carbon credit price, and is the adjustment parameter. Where, The value of can be 0.05, The value of can be 0.03.

[0102] In summary, the system can achieve precise assessment and pricing of green loan risks, while promoting the activity of the carbon market and promoting the development of green finance.

[0103] In summary, the green loan risk assessment and pricing system provided by the present application has the following 5 aspects of beneficial effects compared with the prior art:

[0104] 1. Full industry chain analysis:

[0105] Unlike the existing bank industry, which assesses loan credit based on the historical financial income and expenditure data of the applicant enterprise, the present application provides a green loan exclusive large model. Through the bank green loan private domain information base model training, the full industry chain green production capacity of the industry chain where the applicant enterprise is located is analyzed. Through deep learning algorithm, the green loan information base problem processing, business process, industry information of the industry chain where the applicant enterprise is located, and the key technology of green production capacity form green loan experience accumulation; the model architecture of green loan, financial knowledge concept, industry market logic, professional terms form dynamic knowledge input, realizing unsupervised learning and artificial intelligence automatic updating.

[0106] 2. Full product chain analysis:

[0107] In the current risk rating system of the banking industry, the traditional five-level risk loan classification mainly focuses on single loan business, and its evaluation range is relatively narrow, which is difficult to fully reflect the complexity and potential risks of green financial business. Compared with this single perspective risk assessment method, the present application relies on advanced deep learning algorithm, not only pays attention to the risk characteristics of green loan itself, but also expands the evaluation range to the entire green financial product chain, covering green bonds, green funds, carbon financial special products and green leasing and other diversified financial products. Through the strong learning ability of the deep learning model, the system can deeply mine the internal relationship and mutual influence among these financial products, so as to build a more comprehensive and accurate risk assessment framework.

[0108] In specific implementation, the system first collects and integrates the relevant data of these green financial products, including historical performance, market environment, policy orientation and other dimensions. Then, the deep learning algorithm is used to deeply mine and analyze these data, extract the key risk characteristics, and establish the corresponding risk model. On this basis, the system can realize the comprehensive risk rating of green financial product combination, and provide more comprehensive and detailed risk assessment results for banks. This cross-product line comprehensive risk assessment method not only helps banks to more accurately grasp the overall risk level of green financial business, but also effectively reduces the risk of individual green loan business. Through optimizing resource allocation and strengthening risk management, banks can further improve the efficiency and safety of their green loan business, and contribute to the sustainable development of green finance.

[0109] The present invention is no longer limited to the risk rating of single green loan business, but extends the evaluation range to green bonds, green funds, carbon finance special products, green leases and other green finance full product chains. Through deep learning algorithm, the system can comprehensively consider the risk correlation and mutual influence among various products, realize more accurate and comprehensive risk assessment, and provide more reliable risk management basis for banks.

[0110] 3. Carbon balance and carbon absorption evaluation and carbon absorption enterprise matching:

[0111] The present system provides a more comprehensive carbon balance perspective by considering both carbon emissions and carbon absorption of enterprises. This evaluation method not only focuses on reducing carbon emissions, but also includes the potential of increasing carbon absorption through biological carbon sequestration and carbon capture and storage (CCUS) technology. This enables banks to more accurately assess the eligibility and risk level of green loans for enterprises.

[0112] Using artificial intelligence large models, the present system can actively identify and match potential carbon absorption enterprises that may increase carbon credits for green loan applicants through forest planting or other biological carbon sequestration methods, or through CCUS technology. This matching mechanism optimizes the loan granting process and promotes cooperation within the green finance ecosystem.

[0113] The present system uses deep learning algorithms to dynamically adjust loan pricing based on the carbon balance of enterprises. This method takes into account the value of carbon credits, as well as the impact of market and policy changes on the value of carbon credits, providing banks with a more flexible and responsive market change pricing strategy.

[0114] Combined with deep learning algorithms, the present system can realize professional evaluation and pricing of green loans for full industry chain analysis. This pricing mechanism not only considers the financial situation of enterprises, but also integrates environmental benefits and key technical factors of green production capacity, making loan pricing more scientific and reasonable. Compared with existing technologies, the present system can better encourage enterprises to make green investments and promote sustainable development.

[0115] 4. Deep learning model:

[0116] The present invention uses deep learning algorithms to automatically extract complex and high-dimensional risk features from massive data. These features not only include traditional financial indicators, but also include multi-dimensional information such as environment, society and governance (ESG), making risk assessment more comprehensive and in-depth. At the same time, the deep learning model can handle non-linear relationships and implicit patterns between data, improving the accuracy of risk assessment.

[0117] After implementing the existing green loan large model and depositing private domain information, the deep learning algorithm can realize the automatic interaction information and private domain management of the green loan application enterprise, predict customer behavior, and then realize the professional evaluation and pricing of green loans for the whole industry chain analysis.

[0118] The system analyzes the whole industry chain green production capacity of the applicant enterprise through deep learning algorithm. Compared with the traditional evaluation model which only relies on historical financial data of the enterprise, this method can more comprehensively capture the potential value and risk of the enterprise in green transformation. This analysis can identify the positioning of the enterprise in the green economy and predict its competitiveness and growth in the green industry, thereby providing more accurate loan decision support for banks.

[0119] 5. Model updating and optimization:

[0120] Based on the preliminary application, we continuously collect new loan data and project feedback to continuously optimize the deep learning model. Including adjusting the network structure, increasing the regularization term to prevent overfitting, and introducing more rich features. In particular, we try to apply graph neural networks (GNN) to the correlation analysis between green projects to further improve the accuracy of risk assessment.

[0121] After the model is optimized, it can significantly improve the accuracy of risk assessment and the rationality of pricing. Especially in identifying high-risk loan projects and setting reasonable loan interest rates, the model's prediction results are highly consistent with the actual situation.

[0122] At the same time, we also evaluate the application effect of the system in actual business through customer satisfaction survey, loan default rate and other indicators. Thus, it is conducive to improving the processing efficiency of bank green loan business and reducing the risk of loan default, promoting the sustainable development of green finance.

[0123] The system can form dynamic knowledge input of green loan model architecture, financial knowledge concept, industry market logic, and professional terms, realize unsupervised learning and artificial intelligence automatic update. This advantage enables the system to continuously optimize itself, adapt to market changes, improve the timeliness and accuracy of risk assessment, and greatly reduce the labor cost and time delay compared with traditional models that need to be updated and maintained manually.

[0124] The invention not only provides static risk rating, but also can update risk assessment results in real time according to market changes, policy guidance and other factors, realizing dynamic risk management. This helps banks to adjust green financial product portfolio in time, optimize decision-making process, reduce the risk of individual green loan business, and at the same time improve overall business efficiency.

[0125] In summary, the green loan risk assessment and pricing system provided by the present application realizes accurate assessment and pricing prediction of green loan risks by comprehensively using deep learning algorithms and related data, and provides strong support for the healthy development of bank green loan business.

[0126] The above shows and describes the basic principles and main features of the present application and the advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A green loan risk assessment and pricing system, characterized in that, include: Data collection module: used to collect financial data, supply chain information, and environmental data of companies applying for green loans; collect historical data on green loans and bank green finance products; and collect data related to carbon emissions and carbon absorption of loan companies. Data processing module: Used to preprocess the collected data, including cleaning, noise reduction, removal of outliers and missing values, standardization, and normalization. Feature extraction module: used to extract features from the preprocessed data and construct feature vectors for use as input to subsequent deep learning models; Model training module: used to design a deep learning model for risk assessment and pricing, including convolutional layers and recurrent layers, and to train the model using the feature vectors; Assessment and pricing module: Used to conduct risk assessment on new loan applications, as well as carbon absorption potential assessment and carbon balance risk assessment using a trained model, and to price loans in combination with key technical factors of environmental benefits and green production capacity. Based on the assessment results and loan pricing recommendations output by the model, green loan business strategies are formulated. Model optimization module: used to dynamically update the model according to market changes and policy guidance, and to adjust the loan credit limit and pricing using optimization algorithms based on risk assessment results, as well as to collect carbon emission and carbon absorption data of enterprises after the loan is issued for model feedback and iterative optimization; The carbon absorption enterprise matching module is used to identify and match carbon absorption enterprises in the bank's private information database using artificial intelligence algorithms.

2. The green loan risk assessment and pricing system according to claim 1, characterized in that, The formula used in adjusting the loan credit limit and pricing based on the risk assessment results using the optimization algorithm is shown in the following formula (1): （1） In the formula, For loan interest rates, Based on the interest rate, It is a natural constant. To score the risk, For market carbon credit prices, and To adjust the parameters.

3. The green loan risk assessment and pricing system according to claim 1, characterized in that, The assessment and pricing module specifically includes: pricing loans based on the comprehensive risk rating output by the model, market interest rates, and green project yields, wherein the formula used is shown in equation (2) below: （2） In the formula, Pricing loans, Risk_Rating is the overall risk rating. Market_Rate is the market interest rate. Green_Project_ROI, i.e., the rate of return on green projects. , , , For regression coefficients, This is the error term.

4. The green loan risk assessment and pricing system according to claim 1, characterized in that, The carbon absorption potential assessment includes evaluating the increased carbon absorption potential of enterprises through biocarbon sequestration and CCUS technologies, and the carbon balance risk assessment includes conducting risk assessment using a deep learning model based on the enterprise's carbon emission potential and carbon absorption potential.

5. A green loan risk assessment and pricing system according to claim 4, characterized in that, The formula used in the carbon absorption potential assessment is shown in equation (3) below: T （3） In the formula, For carbon absorption potential, Where is forest cover, and T is the technological carbon sequestration capacity. , , These are model parameters; The formula used in the carbon balance risk assessment is shown in equation (4) below: （4） In the formula, R is the carbon balance risk score, and e is the natural constant. For carbon emissions, For carbon absorption potential, As a policy influencing factor, As a market sensitivity factor, , , , These are the weight parameters.

6. A green loan risk assessment and pricing system according to claim 1, characterized in that... The formula used in the carbon absorption enterprise pairing module is shown in equation (5) below: （5） In the formula, This is the minimum value among all pairing schemes, i.e., the value corresponding to the optimal pairing scheme. The total number of carbon-emitting companies applying for loans. For the i-th carbon-emitting enterprise applying for a loan, For the j-th carbon-absorbing enterprise in the bank's private domain information database, Let be the distance function between the i-th carbon-emitting enterprise applying for a loan and the j-th carbon-absorbing enterprise. Let j be the cost function of the j-th carbon absorbing firm. To weigh the parameters.

7. A green loan risk assessment and pricing system according to claim 1, characterized in that, The model training module includes: introducing a Dropout layer to reduce overfitting, constructing a model by combining an attention mechanism, and optimizing model parameters using a backpropagation algorithm.

Images