Financial data generation method, system and device based on vine copula and storage medium

Abstract

The application provides a financial data generation method, system and device based on vine copula and a storage medium, and comprises the following steps: obtaining financial data, converting the financial data into data conforming to the definition domain of a vine copula model, and saving corresponding conversion rules; determining the structure of the vine copula model and estimating corresponding parameters by using the maximum likelihood estimation method based on the converted financial data; generating model data based on the determined vine copula model; and converting the model data into data consistent with the original data scale by reversely applying the conversion rules, so as to obtain financial data based on vine copula. In order to solve the problems in the prior art, the method is a new data generation technology, and is used for solving the problem that a financial institution cannot establish a model or the effect of the model is poor due to data shortage. The technology is based on a probability statistical model, and can mine the dependency relationship between data features while generating new data.

Description

Technical Field

[0001] This invention belongs to the field of financial data processing technology, and in particular relates to a financial data generation method, system, device and storage medium based on Fujicopula. Background Technology

[0002] With the development of artificial intelligence, the financial industry has gained a deeper understanding of the application value of information technology. Artificial intelligence models are being applied to various aspects of the financial industry, such as helping financial institutions make loan decisions and conducting precise marketing of financial products.

[0003] However, financial institutions often encounter insufficient or nonexistent data when building AI models, leading to problems such as the inability to build models or unstable models. For example, financial institutions with a short history often lack sufficient data accumulation due to their short timeframe. Even for established financial institutions, there may not be enough data to build models due to new products or application scenarios. On the other hand, with the continuous improvement of data security legislation in my country and the increasing standardization of data use, restrictions on the use of external data mean that problems such as insufficient data cannot be solved by introducing external data.

[0004] Therefore, using generated data that closely resembles real data to build an initial model, and then iterating based on actual results during model usage, is a compliant and convenient approach. Generally, there are two common generative data techniques: one is based on deep neural networks, such as generative adversarial networks (GANs); the other is based on statistical probability models to uncover the interdependencies between data points, such as multidimensional normal distribution models.

[0005] There are two common generative data generation techniques: one is based on deep neural networks, such as Generative Adversarial Networks (GANs). GANs use adversarial training to make the generated samples conform to the distribution of real data. In a GAN, two networks are trained adversarially. One is the generator network, whose goal is to generate samples whose source the discriminator network cannot distinguish; the other is the discriminator network, whose goal is to accurately determine whether a sample comes from real data or was generated by the generator network. These two networks with opposing goals are trained alternately. When convergence occurs, if the discriminator network can no longer determine the source of a sample, the generator network can generate samples that conform to the distribution of real data.

[0006] Another data generation technique based on common probability models is more traditional. It involves mining the interdependencies between data points to establish a probability model that best reflects the actual data distribution. This model is then used to generate data that conforms to the model. For example, the multidimensional normal distribution model assumes that the multidimensional data distribution follows a multi-normal distribution. By fitting the relevant parameters, a probability model is established, and then new data is generated.

[0007] Data generation techniques based on deep learning networks share the same drawbacks as deep learning networks themselves. First, training deep learning networks is costly, requiring substantial computational resources. Second, while deep learning networks generally fit data well, their complex structure leads to poor interpretability, making it difficult to verify model correctness and measure dependencies between multidimensional data. However, in the financial industry, interpretability is extremely important; besides good model performance, the model must also be logically constructed and conform to common financial sense.

[0008] On the other hand, while data generation techniques based on common probability models offer better interpretability, these models typically rely on strong pre-existing assumptions. The selection of a particular probability model presupposes that the data distribution conforms to that model. For low-dimensional data, the appropriate probability model can be determined through observation or statistical methods. However, if the data dimensionality is high, it becomes difficult to determine whether the data distribution conforms to the model. For example, the multidimensional normal distribution model mentioned earlier, if used to fit data, assumes that the data distribution conforms to a multidimensional normal distribution. This requires that the marginal distributions of the data all conform to a normal distribution, and that any combination of data also conforms to a normal distribution. However, real-world high-dimensional data rarely meets this condition. Summary of the Invention

[0009] To address the problems of existing technologies, this invention provides a financial data generation method, system, device, and storage medium based on Tengcopula. This method is a novel data generation technology designed to solve the problem of financial institutions being unable to build models or experiencing poor model performance due to data shortages. Based on probabilistic statistical models, this technology can simultaneously generate new data and uncover the dependencies between data features.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] A financial data generation method based on Fuji Copula includes:

[0012] Acquire financial data, transform the financial data into data that conforms to the domain defined by the Tengcopula model, and save the corresponding transformation rules; the financial data includes internal financial data of financial institutions used to build artificial intelligence models and / or external financial data of financial institutions that are associated with financial data.

[0013] Based on the transformed financial data, the structure of the Fujicopula model and the corresponding parameters are determined by the maximum likelihood estimation method.

[0014] Based on the determined vine copula model, model data is generated;

[0015] By applying the transformation rules in reverse, the model data is transformed into data with the same scaling as the original data, resulting in financial data based on Tengcopula.

[0016] As a further improvement of the present invention, the acquisition of financial data, the transformation of the financial data into data conforming to the domain of the Tengcopula model, and the saving of the corresponding transformation rules include:

[0017] Acquire financial data and transform it in different ways according to the type of financial data:

[0018] For continuous financial data, a distribution function F is established, and for each data point x... i Perform data transformation;

[0019] u i =F(x) i )

[0020] For discrete financial data, for each data point x i The calculation is performed using the following method:

[0021]

[0022] Where n represents the number of samples, r i Represents x i The order in which all data is sorted from smallest to largest;

[0023] The transformation makes the domain of financial data fall within the interval [0,1].

[0024] As a further improvement of the present invention, the determination of the structure of the Fujicopula model and the estimation of the corresponding parameters based on the transformed financial data using the maximum likelihood estimation method includes:

[0025] The optimal structure of the vine copula is determined by maximum likelihood estimation, and the connection between each variable is obtained. The optimal copula function is selected from the candidate binary copula functions, and the corresponding parameters of the function are estimated.

[0026] As a further improvement of the present invention, the step of generating model data based on the determined Fujicopula model includes:

[0027] Based on the obtained vine copula model, new data is generated, and... The model generates a dataset where the data is distributed in the range [0,1].

[0028] As a further improvement of the present invention, the reverse application of the transformation rule to transform the model data into data with the same scale as the original data includes:

[0029] For continuous financial data, for each data point, calculate... As generated data values;

[0030] For discrete financial data, first calculate the corresponding data range. Then, based on the previous sorting, find... The corresponding data value is used as the generated data value;

[0031] When all generated data After transformation, the generated dataset is obtained.

[0032] As a further improvement of the present invention, the transformation rule includes the transformation order and the corresponding actual data value, with a one-to-one correspondence between the order and the corresponding actual data value. This applies to all generated data. After transformation, the generated dataset is obtained.

[0033] As a further improvement to the present invention, the process of obtaining financial data based on Fujicopula further includes:

[0034] The obtained financial data based on Fujicopula is sent to the financial institution as input data for the artificial intelligence model.

[0035] A financial data generation system based on Fuji Copula, comprising:

[0036] The data transformation module is used to acquire financial data, transform the financial data into data that conforms to the domain defined by the Tengcopula model, and save the corresponding transformation rules; the financial data includes internal financial data of financial institutions used to build artificial intelligence models and / or external financial data of financial institutions that are related to the financial data.

[0037] The Fujicopula determination module is used to determine the structure of the Fujicopula model and estimate the corresponding parameters based on the transformed financial data using the maximum likelihood estimation method.

[0038] The data generation module is used to generate model data based on the determined vine copula model;

[0039] The reverse transformation module is used to reverse the transformation rules and transform the model data into data with the same scale as the original data, thereby obtaining financial data based on Tengcopula.

[0040] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the Fujicopula-based financial data generation method.

[0041] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the financial data generation method based on Fujicopula.

[0042] Compared with the prior art, the beneficial effects of the present invention are reflected in:

[0043] The data generation technology employed in this invention is based on the vine copula model. Using this data generation technology, the problem of financial institutions being unable to perform artificial intelligence modeling due to a lack of data is addressed. Compared to existing data generation technologies, the vine copula model, which is essentially a probabilistic model, differs from deep learning networks in that it offers excellent interpretability and low computational cost. Furthermore, because the model itself has a tree-like structure composed of multiple binary functions with a wide variety of possible structures, it can flexibly fit data and capture dependencies between data points. Attached Figure Description

[0044] Figure 1 A schematic diagram of an existing vine structure (using a 5-dimensional vine structure as an example);

[0045] Figure 2 The existing standard process for building artificial intelligence models;

[0046] Figure 3 The present invention provides a flowchart of the modeling process in which a data generation module is introduced in an optional embodiment;

[0047] Figure 4 This invention provides a flowchart of the Fujicopula financial data generation technology according to an optional embodiment;

[0048] Figure 5 The present invention provides a histogram of bank income and expenditure data; (a) represents income, and (b) represents expenditure.

[0049] Figure 6 A comparison chart of data distribution for generating datasets based on Fujicopula financial data is shown. (a) and (c) are the original data, and (b) and (d) are the data generated through simulation.

[0050] Figure 7 This is a schematic diagram of the financial data generation system based on Fujicopula according to the present invention;

[0051] Figure 8 This is a schematic diagram of an electronic device structure according to the present invention. Detailed Implementation

[0052] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0053] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0054] Terminology Explanation:

[0055] Marginal distribution: refers to the probability distribution of a multidimensional random variable in probability theory and statistics that contains only a subset of the variables.

[0056] Joint distribution: For two random variables X and Y, the joint distribution is the probability distribution for both X and Y. Similarly, for multidimensional variables, the joint distribution is the probability distribution for all variables.

[0057] Copula function: The Copula function is based on Sklar's theorem, which considers a d-dimensional vector X = (X1, ..., X2) / (X3, ..., X4). d ), whose joint probability density function is f, Sklar's theorem shows that for each marginal distribution F1, ..., F d For the joint distribution F, there exist some d-dimensional Copula functions C: [0, 1] d →[0, 1], satisfying:

[0058] F(x) = C(F1(x1), ..., F d (x d x = (x1, ..., x) d )

[0059] Note: Essentially, the copula function is a joint probability distribution.

[0060] Note: A binary copula function is a copula function of two variables.

[0061] The vine copula function: First, we need to understand the vine structure: A d-dimensional regular vine structure v consists of d-1 trees T1, ..., Td-1 Composition, the node set N of each tree i and edge set E i The following conditions must be met:

[0062] The set of nodes of tree T1 is N1 = {1, ..., d}, and the set of edges is E1;

[0063] When i = 2, ..., d-1, tree T i The node is N i =E i-1 .

[0064] Proximity condition: If tree T i The two edges in tree T i+1 If two edges are connected by an edge, then the two edges must have a common node.

[0065] The vine copula model is a model constructed from multiple binary copula functions based on a conventional vine structure.

[0066] For example: a 5-dimensional vine structure, such as Figure 1 As shown, assigning a binary copula function to the two variables connected by each edge constitutes a vine copula function.

[0067] The typical process for building an artificial intelligence model is as follows:

[0068] 1) Input data and perform data preprocessing;

[0069] 2) Select and train the model based on the requirements and specific scenarios;

[0070] 3) Evaluate the model. If the model performs poorly, it needs to be readjusted; if it performs well, output the model for application. (See below.) Figure 2 As shown.

[0071] Ideally, financial institutions have enough data to train and debug models. However, when data is insufficient, financial institutions need to supplement the data in order to build models.

[0072] Considering the characteristics of financial data, this invention proposes a novel data generation technique. Based on the use of the Teng copula model, it mines the joint relationships between multidimensional data and generates new data for subsequent modeling work. This proposed data generation technique aims to solve the problem of data shortage preventing model building. A data generation module is introduced into the modeling process, as follows: Figure 3 As shown.

[0073] The vine copula model is essentially a probabilistic model, and therefore, unlike deep learning networks, it offers excellent interpretability and low computational cost. Furthermore, because the model itself has a tree-like structure composed of multiple binary functions, and these functions offer a wide variety of possible structures, the model can flexibly fit data and capture dependencies between data points.

[0074] like Figure 4 As shown, the financial data generation technology process based on Fujicopula is as follows:

[0075] Acquire financial data, transform the financial data into data that conforms to the domain defined by the Tengcopula model, and save the corresponding transformation rules; the financial data includes internal financial data of financial institutions used to build artificial intelligence models and / or external financial data of financial institutions that are associated with financial data.

[0076] Based on the transformed financial data, the structure of the Fujicopula model and the corresponding parameters are determined by the maximum likelihood estimation method.

[0077] Based on the determined vine copula model, model data is generated;

[0078] By applying the transformation rules in reverse, the model data is transformed into data with the same scaling as the original data, resulting in financial data based on Tengcopula.

[0079] The specific steps are explained in detail below:

[0080] Step 0: Confirm the available binary copula functions. Generally, you can choose a few common ones that are sufficient to cover the data distribution.

[0081] Step 1: Obtain the financial data needed to build the artificial intelligence model from the financial database of financial institutions, transform the existing data into data that conforms to the domain of the Fujicopula model, and save the corresponding transformation rules. The original financial data is often not in the interval [0,1], while the domain of the Fujicopula model is in [0,1].

[0082] Based on the data type of financial data, different conversion methods are used:

[0083] For continuous financial data (such as deposit amounts), a corresponding marginal cumulative distribution function F is established for each data point x. i Perform data transformation:

[0084] u i =F(x) i )

[0085] Other functions can be used instead of F, such as the empirical cumulative distribution function.

[0086] For discrete financial data (e.g., education level), for each data point x i The following formula can be used for calculation:

[0087]

[0088] Here, n represents the number of samples, and r i Represents x i The order in which all data is sorted from smallest to largest.

[0089] For example, banks need income and expenditure data for modeling, collect relevant external data, and perform preprocessing. Generally, both income and expenditure are higher than 0, and the distribution exhibits a significant right-tailed tendency (e.g., ...). Figure 5 As shown in (a)). Following the above conversion method, the income data is converted to the [0,1] interval (as shown in the diagram). Figure 5 (as shown in (b)).

[0090] Step 2: Based on the transformed data, determine the structure of the Fujicopula model and estimate the corresponding parameters using the maximum likelihood estimation method.

[0091] Maximum likelihood estimation can determine the optimal structure of the copula, i.e., the connection method between the variables. Simultaneously, the optimal copula function can be selected from the candidate binary copula functions, and the corresponding parameters can be estimated.

[0092] Example: Suppose there are 5 variables, represented by 1, 2, 3, 4, and 5. There may be multiple connection relationships between these 5 variables, such as... Figure 1 As shown.

[0093] This step determines the specific structure, as well as the binary copula function and parameters corresponding to each pair of connected variables.

[0094] Step 3: Based on the estimated vine copula model, generate new model data. At this point, the data is distributed in [0,1].

[0095] Based on the vine copula model obtained in step 2, new data can be generated, using... The model generates a dataset where the data is distributed in the range [0,1].

[0096] Step 4: Reverse the data transformation rules of Step 1 to transform the model data into data with the same scale as the original data, and obtain financial data based on Tengcopula.

[0097] For continuous financial data, for each data point, calculate... As generated data values.

[0098] For discrete financial data, the corresponding... Then, based on the previous sorting, find... The corresponding data value is used as the generated data value.

[0099] Example: In step 1, this invention stores the data transformation rules, recording the order and corresponding actual data values. Assume the correspondence is as follows:

[0100]

[0101]

[0102] The next step is simply to determine The range within which the data is located is used to determine the corresponding data value, which is then used as the generated data value. If the calculated... Use a3 as the generated data value.

[0103] When all the data generated in step 3 After the above transformation, the generated dataset is obtained.

[0104] like Figure 6 The image shows a comparison of the data distribution of the financial data generated using the Tengcopula dataset. (a) and (c) are the original data, while (b) and (d) are the data generated through simulation. It can be seen that the distributions in the left and right columns are very similar, indicating that the generated data and the original data have essentially the same distribution. Furthermore, considering the dependency between the two variables (income and expenditure), the correlation coefficient was calculated. The correlation coefficient for the original data is 0.3189, while the correlation coefficient for the simulated data is 0.3131. This demonstrates that the proposed data generation method can effectively capture the dependencies between variables.

[0105] Taking banks as an example, if a bank needs income and expenditure data to build a model (generally used for user profiling), the income and expenditure data of the bank's own accounts can only reflect partial income and expenditure information. This is because the bank cannot determine whether a user has accounts or other deposits at other banks. Therefore, when building an artificial intelligence model, the bank needs more comprehensive income and expenditure information, turning to external data and performing financial data modeling based on Fujicopula. Then, data is generated through simulation using the financial data model based on Fujicopula and migrated to the internal artificial intelligence model for use.

[0106] Therefore, the final financial data obtained based on Fujicopula also includes:

[0107] The obtained financial data based on Fujicopula is sent to the financial institution as input data for the artificial intelligence model, thus initiating the subsequent training process of the AI ​​model.

[0108] like Figure 7 As shown, the present invention also provides a financial data generation system based on Fujicopula, comprising:

[0109] The data transformation module is used to acquire financial data, transform the financial data into data that conforms to the domain defined by the Tengcopula model, and save the corresponding transformation rules; the financial data includes internal financial data of financial institutions used to build artificial intelligence models and / or external financial data of financial institutions that are related to the financial data.

[0110] The Fujicopula determination module is used to determine the structure of the Fujicopula model and estimate the corresponding parameters based on the transformed financial data using the maximum likelihood estimation method.

[0111] The data generation module is used to generate model data based on the determined vine copula model;

[0112] The reverse transformation module is used to reverse the transformation rules and transform the model data into data with the same scale as the original data, thereby obtaining financial data based on Tengcopula.

[0113] like Figure 8 As shown, a third objective of this invention is to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the financial data generation method based on Fujicopula.

[0114] The financial data generation method based on Fujicopula includes:

[0115] Acquire financial data, transform the financial data into data that conforms to the domain of the Tengcopula model, and save the corresponding transformation rules;

[0116] Based on the transformed financial data, the structure of the Fujicopula model and the corresponding parameters are determined by the maximum likelihood estimation method.

[0117] Based on the determined vine copula model, model data is generated;

[0118] By applying the transformation rules in reverse, the model data is transformed into data with the same scaling as the original data, resulting in financial data based on Tengcopula.

[0119] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the financial data generation method based on Fujicopula.

[0120] The financial data generation method based on Fujicopula includes:

[0121] Acquire financial data, transform the financial data into data that conforms to the domain of the Tengcopula model, and save the corresponding transformation rules;

[0122] Based on the transformed financial data, the structure of the Fujicopula model and the corresponding parameters are determined by the maximum likelihood estimation method.

[0123] Based on the determined vine copula model, model data is generated;

[0124] By applying the transformation rules in reverse, the model data is transformed into data with the same scaling as the original data, resulting in financial data based on Tengcopula.

[0125] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0126] 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.

[0127] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0128] 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.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for generating financial data based on Fujicopula, characterized in that, include: Acquire financial data, transform the financial data into data that conforms to the domain defined by the Tengcopula model, and save the corresponding transformation rules; the financial data includes internal financial data of financial institutions used to build artificial intelligence models and / or external financial data of financial institutions that are associated with financial data. Based on the transformed financial data, the structure of the Fujicopula model and the corresponding parameters are determined by the maximum likelihood estimation method. Based on the determined vine copula model, model data is generated; By applying the transformation rules in reverse, the model data is transformed into data with the same scaling as the original data, resulting in financial data based on Tengcopula. The reverse application of the transformation rules to transform the model data into data with the same scale as the original data includes: For continuous financial data, for each data point, calculate... As generated data values; For discrete financial data, first calculate the corresponding data range. Then, based on the previous sorting, find... The corresponding data value is used as the generated data value; When all generated data After transformation, the generated dataset is obtained. ; The transformation rules include the transformation order and the corresponding actual data values, with a one-to-one correspondence between the order and the corresponding actual data values. This applies to all generated data. After transformation, the generated dataset is obtained. ; The process of acquiring financial data involves transforming it into data that conforms to the domain defined by the Teng Copula model and saving the corresponding transformation rules, including: Acquire financial data and transform it in different ways according to the type of financial data: For continuous financial data, establish a distribution function. For each data point Perform data transformation; For discrete financial data, for each data point The calculation is performed using the following method: in, Represents the number of samples. represent The order in which all data is sorted from smallest to largest; The transformation makes the domain of financial data fall within the range Inside; Based on the transformed financial data, the structure of the Fujicopula model and the corresponding parameters are determined using the maximum likelihood estimation method, including: The optimal structure of the vine copula is determined by maximum likelihood estimation, and the connection between each variable is obtained. The optimal copula function is selected from the candidate binary copula functions, and the corresponding parameters of the function are estimated.

2. The method according to claim 1, characterized in that, The process of generating model data based on the determined Fujicopula model includes: Based on the obtained vine copula model, new data is generated, and... The model generates a dataset where the data is distributed as follows: Among them.

3. The method according to claim 1, characterized in that, The process of obtaining financial data based on Fujicopula also includes: The obtained financial data based on Fujicopula is sent to the financial institution as input data for the artificial intelligence model.

4. A financial data generation system based on Fujicopula, implementing the financial data generation method based on Fujicopula as described in any one of claims 1-3, characterized in that, include: The data transformation module is used to acquire financial data, transform the financial data into data that conforms to the domain defined by the Tengcopula model, and save the corresponding transformation rules; the financial data includes internal financial data of financial institutions used to build artificial intelligence models and / or external financial data of financial institutions that are related to the financial data. The Fujicopula determination module is used to determine the structure of the Fujicopula model and estimate the corresponding parameters based on the transformed financial data using the maximum likelihood estimation method. The data generation module is used to generate model data based on the determined vine copula model; The reverse transformation module is used to reverse the transformation rules and transform the model data into data with the same scale as the original data, thereby obtaining financial data based on Tengcopula.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the financial data generation method based on Fujicopula as described in any one of claims 1-3.

6. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the financial data generation method based on Fujicopula as described in any one of claims 1-3.

Images