Big data-based digital financial risk management system and management method

Abstract

The application provides a big data-based digital financial risk management system and method, which comprises the following parts: a data collection module, a time domain conversion module, a credit evaluation module, a data slicing module, a weight calculation module, a model optimization module, a resource allocation module and a financial database; the application solves the problem of long processing time and low efficiency caused by large data volume in the prior art when processing big data; the simple linear data conversion may not be sufficient when capturing complex relationships in financial data, resulting in inaccurate or in-depth evaluation; the application realizes a risk evaluation method based on big data and machine learning, which helps to more accurately and efficiently evaluate the credit risk of users and provides important technical support for risk management of financial institutions.

Description

Technical Field

[0001] This invention relates to the field of digital financial technology, and in particular to a digital financial risk management system and method based on big data. Background Technology

[0002] With technological advancements, the volume of data in the financial sector has exploded. These data sources include transaction records, user behavior, market news, and social media feedback, providing financial institutions with unprecedented data resources. Simultaneously, the financial market itself is highly dynamic, influenced by numerous micro and macro factors, which increases the demand for real-time data processing and analysis.

[0003] However, traditional financial risk management methods face many challenges when dealing with large-scale data. For example, they may suffer from performance bottlenecks, employ overly simplistic data transformation and representation methods, or lack in-depth evaluation. Furthermore, financial decisions require a high degree of accuracy, as even a small misjudgment can lead to substantial economic losses. Moreover, in the financial sector, particularly in areas like high-frequency trading, real-time data processing has become a critical requirement.

[0004] Chinese patent application number CN202310923664.4, publication date: October 10, 2023, discloses a method for assessing the liquidity risk of commercial banks based on digital finance. The method includes: acquiring digital financial business data, acquiring liquidity risk assessment indicators for commercial banks, acquiring liquidity positions in future periods, and classifying liquidity risk levels and conducting risk assessments. This invention ensures data quality and accuracy by collecting and preprocessing digital financial business data from commercial banks, facilitating subsequent data processing and analysis. Furthermore, it acquires various liquidity risk assessment indicators for commercial banks based on digital finance and obtains their liquidity positions in future periods based on their short-term cash flow situation. This allows for the assessment of the future liquidity risk level of commercial banks based on pre-classified risk levels. Compared to traditional liquidity risk assessment methods, this method offers higher timeliness, more comprehensive assessment criteria, and improved assessment efficiency and accuracy.

[0005] However, in the process of implementing the inventive technical solutions in the embodiments of this application, the inventors of this application have discovered that the above-mentioned technology has at least the following technical problems: the existing technology has long processing time and low efficiency when processing big data due to the huge amount of data; it relies on simple linear data transformation, which may be insufficient when capturing complex relationships in financial data, resulting in assessments that may not be accurate or in-depth enough; many relationships in financial data are high-dimensional, and simple assessment methods may not be able to capture these high-dimensional relationships; it does not consider how to dynamically allocate computing resources according to the characteristics or importance of the data, resulting in low efficiency in digital financial risk management. Summary of the Invention

[0006] This application provides a big data-based digital financial risk management system and method, addressing the shortcomings of existing technologies in handling large datasets. These shortcomings include: long processing times and low efficiency due to the sheer volume of data; reliance on simple linear data transformations, which may be insufficient for capturing complex relationships within financial data, leading to inaccurate or superficial assessments; the high-dimensionality of relationships within many financial data sets, which simple assessment methods may fail to capture; and the lack of consideration for dynamically allocating computing resources based on data characteristics or importance, resulting in inefficient digital financial risk management. The application implements a risk assessment method based on big data and machine learning, which helps to more accurately and efficiently assess users' credit risk, providing crucial technical support for financial institutions' risk management.

[0007] This application provides a digital financial risk management system and method based on big data, specifically including the following technical solutions: A big data-based digital financial risk management system includes the following components: Data collection module, time domain transformation module, credit assessment module, data sharding module, weight calculation module, model optimization module, resource allocation module, financial database; The time-domain transformation module is used to capture short-term and long-term changes and nonlinear relationships in the data. It uses a transformation function to perform nonlinear transformation on the collected data. The time-domain transformation module is connected to the credit assessment module through data transmission. The credit assessment module is used to construct a credit assessment neural network model. The module starts from the input layer, goes through the dimension expansion layer, the mapping layer, the scoring layer, and finally obtains the user's credit score at the output layer. The credit assessment module is connected to the financial database through data transmission. The data sharding module is used to divide big data into small data shards for parallel processing in a cloud computing environment. The data sharding module is connected to the credit assessment module and the weight calculation module through data transmission. The weight calculation module is used to ensure that data slices with large amounts of information receive more computing resources. It uses the concepts of entropy and information gain to define the weight of the data slices. The weight calculation module is connected to the model optimization module through data transmission. The model optimization module is used to define a local error function and a local gradient for each data slice using a distributed gradient descent method, and then calculate the global gradient to optimize and update the model parameters. The model optimization module is connected to the credit assessment module and the resource allocation module through data transmission. The resource allocation module is used to define a resource demand function that takes into account data density, data slice size and processing time, and dynamically allocates computing resources. The resource allocation module is connected to the credit assessment module through data transmission.

[0008] Preferably, it includes the following steps: S100: Acquire user financial data and perform data preprocessing and time-domain transformation; S200: Construct a neural network model for credit assessment, including an input layer, a dimension expansion layer, a mapping layer, a scoring layer, and an output layer; S300: It performs sharding processing on user financial data, dividing big data into small data shards and distributing them to different computing nodes; S400: In terms of parallel computing, it develops an advanced framework for digital financial risk assessment in cloud computing environments, dynamically allocating computing resources.

[0009] Preferably, step S100 specifically includes: A nonlinear transformation function is used to capture short-term and long-term changes and nonlinear relationships in the data. The nonlinear transformation function is defined as follows: , in, It is a non-linear transformation function that transforms the original data Convert to a new representation; This indicates the data that has been retrieved; , , , , These are model parameters.

[0010] Preferably, step S200 specifically includes: The credit assessment neural network model takes preprocessed and transformed data as input to the neural network model, trains the neural network, and finally outputs the user's credit score.

[0011] Preferably, step S200 specifically includes: The input layer passes the data to the dimension expansion layer. In order to further extract the characteristics of the data, the dimension expansion layer introduces data dimension expansion and spatial transformation techniques.

[0012] Preferably, step S200 specifically includes: The dimensional expansion layer passes the expanded data to the mapping layer, which provides a complex nonlinear mapping for the financial data, mapping the dimensionally expanded data to the core risk assessment space and calculating the core risk score. The mapping layer then passes the core risk score to the scoring layer, which provides the user's risk score. Through the construction of a multi-layered credit assessment neural network model, a deep and accurate assessment of the user's credit risk is achieved.

[0013] Preferably, step S300 specifically includes: The acquired user financial data is segmented into smaller data slices and distributed to different computing nodes. Kernel density estimation is used to determine the non-uniformity coefficient of the data, thereby more accurately dividing the data and ensuring that each data slice contains sufficient information.

[0014] Preferably, step S300 specifically includes: Define a data sharding function, and determine the number of data shards based on the sharding function. Determine the partition boundaries and use them to divide the dataset into partitions. Each data subset is sent to a computing node in a cloud computing environment for processing, and each node independently applies a credit assessment neural network model to its assigned data subset.

[0015] Preferably, step S400 specifically includes: In terms of parallel computing, the concepts of entropy and information gain are introduced to define weights, ensuring that data slices with larger amounts of information receive more computing resources, thereby improving overall computing efficiency. Considering the training of the credit assessment neural network model, a distributed gradient descent method was used. A local error function and a local gradient were defined for each data slice, and the global gradient was defined as a weighted average of all local gradients.

[0016] Preferably, step S400 specifically includes: To dynamically allocate computing resources, a resource requirement function is defined that takes into account not only data density, but also the size of data slices and processing time.

[0017] Beneficial effects include: The multiple technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. By introducing machine learning models and combining various mathematical functions and algorithms, such as nonlinear transformation, dimensional expansion, and core risk assessment, we can deeply analyze users' financial data, capture short-term and long-term changes in the data, as well as possible nonlinear and high-dimensional relationships, thereby providing financial institutions with more accurate user credit risk assessment; in addition to traditional transaction records and income information, we also introduce user behavioral data and social network data to more comprehensively assess user credit risk. 2. Parallel computing using cloud computing technology significantly shortens data processing time; data sharding and kernel density estimation techniques ensure fast, uniform, and efficient processing of large datasets; computing resources are dynamically allocated based on the amount of information and processing requirements, ensuring computational efficiency; the weight calculation module also considers entropy and information gain, ensuring that data shards with large amounts of information receive more computing resources. 3. By using loss functions and distributed gradient descent methods, model parameters can be better optimized, ensuring the accuracy of model predictions; by adopting cloud computing and parallel processing technologies, not only is the data processing speed improved, but the cost of processing large-scale data is also effectively reduced.

[0018] 4. The technical solution of this application effectively addresses the shortcomings of existing technologies in handling big data. These include long processing times and low efficiency due to the sheer volume of data; reliance on simple linear data transformation, which may be insufficient for capturing complex relationships in financial data, leading to potentially inaccurate or superficial assessments; the high-dimensionality of relationships within many financial data sets, which simple assessment methods may fail to capture; and the lack of consideration for dynamically allocating computational resources based on data characteristics or importance, resulting in inefficient digital financial risk management. Furthermore, the aforementioned system or method has undergone a series of effectiveness studies and verifications, ultimately enabling a risk assessment method based on big data and machine learning. This helps to more accurately and efficiently assess users' credit risk, providing crucial technical support for financial institutions' risk management. Attached Figure Description

[0019] Figure 1 This is a structural diagram of the big data-based digital financial risk management system described in this application; Figure 2 This is a flowchart of the big data-based digital financial risk management method described in this application; Detailed Implementation

[0020] This application provides a digital financial risk management system and method based on big data, which solves the problems of long processing time and low efficiency caused by the massive amount of data in existing technologies; reliance on simple linear data transformation, which may be insufficient to capture the complex relationships in financial data, resulting in inaccurate or in-depth assessments; the existence of high-dimensional relationships in many financial data, which simple assessment methods may not be able to capture; and the lack of consideration for how to dynamically allocate computing resources according to the characteristics or importance of the data, leading to low efficiency in digital financial risk management.

[0021] The technical solution in this application is to solve the above problems, and the overall approach is as follows: By introducing machine learning models and combining various mathematical functions and algorithms, such as nonlinear transformations, dimensional expansion, and core risk assessment, this approach deeply analyzes users' financial data, capturing short-term and long-term changes and potential nonlinear and high-dimensional relationships within the data. This provides financial institutions with more accurate user credit risk assessments. In addition to traditional transaction records and income information, it incorporates user behavioral data and social network data for a more comprehensive assessment of user credit risk. Utilizing cloud computing technology for parallel computing significantly shortens data processing time. Data sharding and kernel density estimation techniques ensure fast, uniform, and efficient processing of large datasets. Dynamic allocation of computing resources based on data information volume and processing needs ensures computational efficiency. The weight calculation module also considers entropy and information gain, ensuring that data shards with large information volumes receive more computing resources. The use of loss functions and distributed gradient descent methods allows for better optimization of model parameters, ensuring predictive accuracy. The adoption of cloud computing and parallel processing technologies not only improves data processing speed but also effectively reduces the cost of processing large-scale data.

[0022] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0023] See attached document Figure 1 The big data-based digital financial risk management system and method described in this application include the following parts: Data collection module 10, time domain transformation module 20, credit assessment module 30, data sharding module 40, weight calculation module 50, model optimization module 60, resource allocation module 70, financial database 80 The data collection module 10 is used to collect users' financial data, such as transaction records and income information. This data is the basis for assessing users' credit risk. The data collection module 10 is connected to the time domain conversion module 20, the data sharding module 40, and the financial database 80 through data transmission. The time domain transformation module 20 is used to capture short-term and long-term changes in the data, as well as possible nonlinear relationships. It uses a transformation function to perform nonlinear transformation on the collected data. The time domain transformation module 20 is connected to the credit assessment module 30 through data transmission. The credit assessment module 30 is used to construct a credit assessment neural network model. The module starts from the input layer, goes through the dimension expansion layer, the mapping layer, the scoring layer, and finally obtains the user's credit score at the output layer. The credit assessment module 30 is connected to the financial database 80 through data transmission. The data sharding module 40 is used to divide big data into small data shards so that they can be processed in parallel in a cloud computing environment. The data sharding module 40 is connected to the credit assessment module 30 and the weight calculation module 50 through data transmission. The weight calculation module 50 is used to ensure that data slices with large amounts of information receive more computing resources. It uses the concepts of entropy and information gain to define the weight of the data slices. The weight calculation module 50 is connected to the model optimization module 60 through data transmission. The model optimization module 60 is used to define a local error function and a local gradient for each data slice using a distributed gradient descent method, and then calculate the global gradient to optimize and update the parameters of the nonlinear transformation function. The model optimization module 60 is connected to the credit assessment module 30 and the resource allocation module 70 through data transmission. The resource allocation module 70 is used to define a resource demand function, which takes into account data density, data slice size and processing time, and dynamically allocates computing resources. The resource allocation module 70 is connected to the credit assessment module 30 through data transmission. The financial database 80 is used to store user financial data and credit assessment results.

[0024] See attached document Figure 1 The big data-based digital financial risk management method described in this application includes the following steps: S100: Acquire user financial data and perform data preprocessing and time-domain transformation; In the field of digital finance, credit risk assessment has always been an important but challenging task. Existing technologies are often constrained by insufficient data, oversimplification, or human-defined rules. This application proposes a method for assessing users' credit risk using machine learning models. By analyzing users' transaction records, income, and other information, the machine learning model can generate accurate credit scores. The machine learning-based neural network model for credit assessment not only considers users' basic information and transaction records but also incorporates users' behavioral data and social network data to more comprehensively assess users' credit risk.

[0025] The data collection module 10 acquires users' financial data such as transaction records and income, and preprocesses the data. Existing technologies for data preprocessing are relatively mature and can be used. Digital financial data has its own unique characteristics; for example, transaction data may be affected by seasonality, holidays, or special events. Therefore, simple linear transformations are insufficient to capture these details. The time-domain transformation module 20 employs a nonlinear transformation function to capture short-term and long-term changes in the data, as well as possible nonlinear relationships.

[0026] Considering the characteristics of the data, the following nonlinear transformation function is defined: , in, It is a non-linear transformation function that transforms the original data This is transformed into a new representation that can capture short-term and long-term variations in the data, as well as potential non-linear relationships. This refers to the data obtained, such as transaction records, revenue, and other information. , , , , These are the parameters of the nonlinear transformation function, used for the calculation of each part of the nonlinear transformation function.

[0027] S200: Construct a neural network model for credit assessment, including an input layer, a dimension expansion layer, a mapping layer, a scoring layer, and an output layer; The credit assessment module 30 constructs a credit assessment neural network model. Preprocessed and transformed data is used as input to the neural network model. After training, the model outputs the user's credit score. The credit assessment neural network model includes an input layer, a dimension expansion layer, a mapping layer, a scoring layer, and an output layer.

[0028] The input layer passes data to the dimension expansion layer. To further extract the characteristics of the data, the dimension expansion layer introduces data dimension expansion and spatial transformation techniques. Dimension expansion is used to better represent the high-dimensional relationships that may exist in the data. The dimension expansion function is defined as follows: , in, It is a dimension expansion function that transforms the data. The purpose of further transforming it into a high-dimensional space is to capture potential high-dimensional relationships within the data. It is a regularization parameter used to adjust... The degree of curvature of the function ensures It will not excessively distort the data. The specific formula is: , The dimensional expansion layer passes the expanded data to the mapping layer, which provides a complex nonlinear mapping for the financial data, mapping the dimensionally expanded data to the core risk assessment space and calculating the core risk score, as shown in the following formula: , in, It is the core risk score, which will The high-dimensional representation maps to a single risk score. , , , , These are all weight parameters; together they determine how to extract... Risk information is extracted from the high-dimensional representation.

[0029] The mapping layer passes the core risk score to the scoring layer, which then provides the user's risk score. The scoring layer proposes the following scoring function, which combines all the outputs from the previous layers: , in, It is a user risk score. Through the construction of a multi-level credit assessment neural network model, it achieves in-depth and accurate assessment of user credit risk.

[0030] To ensure the accuracy of the model's predictions, a loss function is introduced. This is used to minimize the error between the risk score predicted by the model and the actual risk. The formula for the loss function is: , in, It provides accurate risk labels. The model parameters are adjusted based on the loss function value until the model training is complete, thus providing accurate and stable risk assessment results and offering financial institutions a powerful risk management tool.

[0031] S300: It performs sharding processing on user financial data, dividing big data into small data shards and distributing them to different computing nodes; User credit risk assessment requires extensive data processing and computation, and traditional single-machine or small-scale cluster environments are insufficient to handle the demands of big data. This application utilizes cloud computing technology, which significantly reduces data processing time and costs through parallel computing.

[0032] In a cloud computing environment, to fully utilize resources, the data sharding module 40 shards the acquired user financial data, dividing the large dataset into smaller data shards and distributing them across different computing nodes. A kernel density estimation method is used to determine the non-uniformity coefficient of the data, thereby more accurately partitioning the data and ensuring that each data shard contains sufficient information. Specifically, this non-uniformity coefficient is calculated using the following formula: , in, Indicates the non-uniformity coefficient. It is the acquired user financial dataset. It is a dataset any point in, The actual distribution function of the data. For kernel function, It's bandwidth. It is the number of data points. It is a dataset The first in Data points, Furthermore, the data sharding function is defined as: , in, It is the data sharding function. It is the density function of the data. It is a predefined constant used to control the sensitivity of the slice function. (Based on the slice function...) Determine the number of data shards The formula is: , in, It is a constant. This indicates rounding up. The formula for determining the partition boundaries is: , in, It is the first The end point of each fragment, Use shard boundaries to divide the dataset. Divided into Each data subset is sent to a computing node in a cloud computing environment for processing, and each node independently applies a credit assessment neural network model to its assigned data subset.

[0033] S400: In terms of parallel computing, develop an advanced framework for digital financial risk assessment in a cloud computing environment, dynamically allocating computing resources.

[0034] In parallel computing, to rationally allocate computing resources, the weight calculation module 50 introduces the concepts of entropy and information gain to define weights. This ensures that data slices with larger amounts of information receive more computing resources, thereby improving overall computing efficiency. The formula for calculating weights is: , in, It is the first The weight of each data slice, Representative at the The first data slice The probability of a class.

[0035] Considering the training of the credit assessment neural network model, a distributed gradient descent method was used. The model optimization module 60 defines a local error function and a local gradient for each data slice, while the global gradient is defined as a weighted average of all local gradients. The specific calculation formula is as follows: , , in, Indicates the first Local error function of each data slice Indicates the first Local gradients of each data slice and They are the first The actual output and predicted output for each data point.

[0036] Furthermore, since multiple model parameter sets exist across multiple cloud computing nodes, these parameter sets need to be merged into a global model. This fusion process employs a gradient-based method. The specific formula is as follows: , in, Represents the global gradient. It is the first Model parameters for each data slice, This represents the weighted sum of local gradients.

[0037] Finally, to dynamically allocate computing resources, the resource allocation module 70 defines a resource demand function. This function considers not only data density but also the size of data slices and processing time. The specific formula is as follows: , in, Indicates the first Resource requirement function for each data slice , , These are weighting coefficients used to adjust the importance of different parts of the resource demand function. Indicates the first Data density of each data slice Indicates the first Processing time for each data slice Indicates the first The processing speed of each data slice It refers to a time interval.

[0038] Through the methods and calculations described above, a high-level framework for digital financial risk assessment in a cloud computing environment has been successfully developed. This framework not only handles large-scale data, ensuring the accuracy and efficiency of the assessment, but also significantly reduces data processing time and costs through parallel computing, providing a powerful tool for digital financial risk management.

[0039] In summary, the big data-based digital financial risk management system and management method described in this application have been completed.

[0040] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: 1. By introducing machine learning models and combining various mathematical functions and algorithms, such as nonlinear transformation, dimensional expansion, and core risk assessment, we can deeply analyze users' financial data, capture short-term and long-term changes in the data, as well as possible nonlinear and high-dimensional relationships, thereby providing financial institutions with more accurate user credit risk assessment; in addition to traditional transaction records and income information, we also introduce user behavioral data and social network data to more comprehensively assess user credit risk. 2. Parallel computing using cloud computing technology significantly shortens data processing time; data sharding and kernel density estimation techniques ensure fast, uniform, and efficient processing of large datasets; computing resources are dynamically allocated based on the amount of information and processing requirements, ensuring computational efficiency; the weight calculation module also considers entropy and information gain, ensuring that data shards with large amounts of information receive more computing resources. 3. By using loss functions and distributed gradient descent methods, model parameters can be better optimized, ensuring the accuracy of model predictions; by adopting cloud computing and parallel processing technologies, not only is the data processing speed improved, but the cost of processing large-scale data is also effectively reduced.

[0041] Results Survey: The technical solution of this application effectively addresses the shortcomings of existing technologies in handling big data. These include long processing times and low efficiency due to the sheer volume of data; reliance on simple linear data transformation, which may be insufficient for capturing complex relationships in financial data, leading to potentially inaccurate or superficial assessments; the high-dimensionality of relationships within many financial data sets, which simple assessment methods may fail to capture; and the lack of consideration for dynamically allocating computational resources based on data characteristics or importance, resulting in inefficient digital financial risk management. Furthermore, the aforementioned system or method has undergone a series of effectiveness studies and verifications, ultimately enabling a risk assessment method based on big data and machine learning. This method helps to more accurately and efficiently assess users' credit risk, providing crucial technical support for financial institutions' risk management.

[0042] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0043] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0044] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0045] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A digital financial risk management system based on big data, characterized in that: Includes the following parts: Data collection module, time domain transformation module, credit assessment module, data sharding module, weight calculation module, model optimization module, resource allocation module, financial database; The time-domain transformation module is used to capture short-term and long-term changes and nonlinear relationships in the data. It uses a transformation function to perform nonlinear transformation on the collected data. The time-domain transformation module is connected to the credit assessment module through data transmission. The credit assessment module is used to construct a credit assessment neural network model. This credit assessment neural network model starts from the input layer, goes through the dimension expansion layer, the mapping layer, and the scoring layer, and finally obtains the user's credit score at the output layer. The credit assessment module is connected to the financial database through data transmission. The data sharding module is used to divide big data into small data shards for parallel processing in a cloud computing environment. The data sharding module is connected to the credit assessment module and the weight calculation module through data transmission. The weight calculation module is used to ensure that data slices with large amounts of information receive more computing resources. It uses the concepts of entropy and information gain to define the weight of the data slices. The weight calculation module is connected to the model optimization module through data transmission. The model optimization module is used to define a local error function and a local gradient for each data slice using a distributed gradient descent method, and then calculate the global gradient to optimize and update the parameters of the nonlinear transformation function. The model optimization module is connected to the credit assessment module and the resource allocation module through data transmission. The resource allocation module is used to define a resource demand function that takes into account data density, data slice size and processing time, and dynamically allocates computing resources. The resource allocation module is connected to the credit assessment module through data transmission.

2. The management method of the big data-based digital financial risk management system as described in claim 1, characterized in that, Includes the following steps: S100: Acquire user financial data, and perform data preprocessing and time-domain transformation; S200: Construct a neural network model for credit assessment, including an input layer, a dimension expansion layer, a mapping layer, a scoring layer, and an output layer; S300: It performs sharding processing on user financial data, dividing big data into small data shards and distributing them to different computing nodes; S400: In terms of parallel computing, the concepts of entropy and information gain are introduced to define weights, ensuring that data slices with larger amounts of information receive more computing resources, thereby improving overall computing efficiency. Considering the training of the credit assessment neural network model, a distributed gradient descent method was used. A local error function and a local gradient were defined for each data slice, and the global gradient was defined as the weighted average of all local gradients. In order to dynamically allocate computing resources, a resource requirement function was defined, which not only considers the data density, but also the size of the data slice and the processing time. Through the above process, a high-level framework for digital financial risk assessment in a cloud computing environment was successfully developed, which dynamically allocates computing resources.

3. The management method of the big data-based digital financial risk management system as described in claim 2, characterized in that, Step S200 specifically includes: The credit assessment neural network model takes preprocessed and transformed data as input to the neural network model, trains the neural network, and finally outputs the user's credit score.

4. The management method of the big data-based digital financial risk management system as described in claim 2, characterized in that, Step S200 specifically includes: The input layer passes the data to the dimension expansion layer. In order to further extract the characteristics of the data, the dimension expansion layer introduces data dimension expansion and spatial transformation techniques.

5. The management method of the big data-based digital financial risk management system as described in claim 4, characterized in that, Step S200 specifically includes: The dimensional expansion layer passes the expanded data to the mapping layer, which provides a complex nonlinear mapping for the financial data, mapping the dimensionally expanded data to the core risk assessment space and calculating the core risk score. The mapping layer then passes the core risk score to the scoring layer, which provides the user's risk score. Through the construction of a multi-layered credit assessment neural network model, a deep and accurate assessment of the user's credit risk is achieved.

6. The management method of the big data-based digital financial risk management system as described in claim 2, characterized in that, Step S300 specifically includes: The acquired user financial data is segmented into smaller data slices and distributed to different computing nodes. Kernel density estimation is used to determine the non-uniformity coefficient of the data, thereby more accurately dividing the data and ensuring that each data slice contains sufficient information.

7. The management method of the big data-based digital financial risk management system as described in claim 6, characterized in that, Step S300 specifically includes: Define a data sharding function, and determine the number of data shards based on the sharding function. Determine the partition boundaries and use them to divide the dataset into partitions. Each data subset is sent to a computing node in the cloud computing environment for processing, and each node is independently applied to the data subset allocated by the credit assessment neural network model.

Images