Multi-party data collaboration and credit assessment methods and systems for supply chain finance

Abstract

This invention provides a method and system for multi-party data collaboration and credit assessment in supply chain finance, relating to the field of credit assessment technology. The method includes creating digital identities, constructing a distributed data collaboration mechanism, collecting supply chain finance data through privacy-preserving computation programs, building a supply chain credit assessment network based on business relationships, calculating credit transmission strength to generate dynamic credit characteristics, and calculating credit assessment index values. This invention achieves secure collaborative sharing of multi-party data, protects the privacy of original data, and improves the accuracy and credibility of credit assessments for supply chain participants.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of credit evaluation, in particular to a multi-party data collaboration and credit evaluation method and system for supply chain finance. BACKGROUND

[0002] Supply chain finance is a financial service mode connecting upstream and downstream enterprises in the supply chain. Through the integrated management of fund flow, information flow and logistics in the supply chain, it provides financing, settlement and risk management services for supply chain participants. With the development of digital economy, supply chain finance has become an important way to solve the financing difficulties and high financing costs of small and medium-sized enterprises. Traditional supply chain finance mainly relies on the credit guarantee of core enterprises, and with the development of technology, multi-party data collaboration and credit evaluation methods based on blockchain, big data and artificial intelligence have gradually become a research hotspot in the industry.

[0003] At present, the multi-party data collaboration and credit evaluation in the field of supply chain finance has the following defects and deficiencies:

[0004] There is serious information asymmetry between supply chain participants, and the data of each party is scattered and closed, making it difficult to form a comprehensive credit profile. Financial institutions cannot obtain complete supply chain transaction data, leading to inaccurate risk assessment and low efficiency of financing decision-making, which restricts the healthy development of supply chain finance business.

[0005] Existing data sharing mechanisms often face challenges in data security and privacy protection. Supply chain participants are reluctant to share raw data due to business secret protection considerations, and traditional data collaboration methods cannot effectively balance the contradiction between data value mining and privacy protection, hindering the deep collaboration of multi-party data. SUMMARY

[0006] The multi-party data collaboration and credit evaluation method and system for supply chain finance provided by the embodiments of the present application can solve the problems in the prior art.

[0007] In a first aspect, the multi-party data collaboration and credit evaluation method for supply chain finance comprises:

[0008] Receiving a data collaboration request submitted by a supply chain participant, and creating a digital identity for each supply chain participant in the blockchain network according to the subject information in the data collaboration request;

[0009] Building a distributed data collaboration mechanism based on the digital identity, connecting the local data nodes of the supply chain participants to the blockchain network, deploying smart contracts and privacy computing programs on each local data node, and the privacy computing program performs encrypted calculation on the local data through a federated learning model, only submits the calculation result to the blockchain network, and ensures that the original data does not leave the local node;

[0010] collecting, by the privacy computing program, supply chain finance data from each local data node, extracting a business association relationship between supply chain participants based on the supply chain finance data, constructing a supply chain credit evaluation network according to the business association relationship, taking the supply chain participants as network nodes and taking the business association relationship as network connection edges, calculating a credit conduction strength between the network nodes based on a graph neural network, the credit conduction strength being used to represent a credit influence degree between the supply chain participants;

[0011] generating a dynamic credit feature for each network node according to the credit conduction strength, and calculating a credit evaluation index value of the supply chain participant by weighting using the dynamic credit feature.

[0012] receiving a data collaboration request submitted by a supply chain participant, and creating a digital identity for each supply chain participant in a blockchain network according to subject information in the data collaboration request, including:

[0013] The data collaboration request includes enterprise registration information, business qualification information, legal person information, and data use range;

[0014] The enterprise registration information is compared with an enterprise credit database, the business qualification information is verified with an administrative permission database, and the legal person information is checked with an identity authentication system; a subject trustworthiness score is calculated based on the verification result; and a permission level of the supply chain participant is determined according to the subject trustworthiness score;

[0015] deploying an identity management smart contract in the blockchain network, generating an identity string containing a participant identifier, a permission level identifier, and a time limit identifier for the supply chain participant based on the permission level;

[0016] The identity string is processed by fragmentation to obtain a plurality of identity fragments, and the identity fragments are sent to a preset number of trust nodes in the blockchain network; the trust nodes sign the received identity fragments; the signature results of the trust nodes are collected, and the signed identity fragments are combined to generate a digital identity certificate.

[0017] Based on the digital identity, a distributed data collaboration mechanism is constructed, the local data nodes of the supply chain participants are connected to the blockchain network, and smart contracts and privacy computing programs are deployed at each local data node, the privacy computing program performs encrypted calculation on the local data through a federated learning model, and only the calculation results are submitted to the blockchain network, including:

[0018] receiving a data node access request of a supply chain participant, and assigning a node identifier to the data node based on the digital identity;

[0019] A three-layer smart contract architecture is deployed in the data node, including a data access layer contract for data access control, a business logic layer contract for feature extraction, and a collaborative computing layer contract for computing task distribution; the data access layer contract verifies the data operation authority according to the node identifier;

[0020] The local data of the data node is preprocessed and feature-extracted by the business logic layer contract to generate a training sample set; a federated learning model is constructed based on the training sample set, and a homomorphic encryption algorithm is used to encrypt the model parameters;

[0021] The encrypted model parameters are distributed to each data node through the collaborative computing layer contract, each data node calculates the model gradient value based on the training sample set, and adds differential privacy noise based on the Laplace distribution to the model gradient value; the intensity of the differential privacy noise is adaptively adjusted according to the calculation round;

[0022] The model gradient value after adding differential privacy noise is aggregated in a weighted average manner to generate a new round of model parameters; the new round of model parameters are cross-validated among the data nodes through the data access layer contract, and the new round of model parameters are written into the blockchain network after validation.

[0023] The model gradient value is added with differential privacy noise based on the Laplace distribution; the intensity of the differential privacy noise is adaptively adjusted according to the calculation round, which includes:

[0024] The value range of each parameter in the model gradient value is calculated, the sensitivity of the model gradient value is determined based on the value range, and the sensitivity represents the maximum influence degree of a single sample on the model gradient value;

[0025] An initial privacy budget and a budget decay coefficient are set, the initial privacy budget is divided by the sensitivity to obtain an initial scale parameter of the Laplace distribution, and the location parameter of the Laplace distribution is determined by the model gradient value;

[0026] The probability density function of the Laplace distribution is constructed based on the initial scale parameter and the location parameter, differential privacy noise is sampled from the probability density function; the differential privacy noise is superimposed on the model gradient value to generate a noisy gradient value that satisfies differential privacy protection;

[0027] The initial privacy budget is multiplied by the training round power of the budget decay coefficient to obtain the privacy budget of the current round; the privacy budget of the current round is divided by the sensitivity to obtain the scale parameter of the Laplace distribution of the current round;

[0028] The updated Laplace distribution is used to protect the differential privacy of the model gradient value of the next round.

[0029] A supply chain credit assessment network is constructed based on the aforementioned business relationships, with supply chain participants as network nodes and business relationships as network connecting edges. The credit transmission strength between network nodes is calculated based on a graph neural network, including:

[0030] Based on the business relationships, a multi-layered information propagation structure for the supply chain credit transmission network is constructed. In each layer of information propagation, a message vector between node pairs is calculated based on the node feature vectors corresponding to the network nodes. The message vector is obtained by performing a linear transformation on the node feature vectors and then passing it through a non-linear activation function. The message vector is multiplied by the association weight between the node pairs to obtain a weighted message vector.

[0031] An attention mechanism is introduced to selectively aggregate the neighbor information of network nodes, calculate the attention weight between the network node and its neighbor nodes, multiply the attention weight with the weighted message vector and sum them to obtain the propagation characteristics of the network node in the current layer.

[0032] The propagation characteristics of network nodes are updated iteratively in multiple layers. In each layer of information propagation, message vector calculation and attention aggregation operations are repeatedly performed to obtain the final feature representation of the network nodes.

[0033] The credit transmission strength between node pairs is calculated based on the final feature representation. The credit transmission strength is obtained by calculating the similarity of node features and then normalizing it to obtain the final transmission probability.

[0034] Based on the credit transmission strength, dynamic credit characteristics are generated for each network node. These dynamic credit characteristics are then used to calculate weighted credit assessment index values ​​for supply chain participants, including:

[0035] Time information is encoded using trigonometric functions of different frequencies to obtain a time encoding vector. The time encoding vector and historical credit records are then input into a long short-term memory network to generate a memory state vector for the network nodes.

[0036] The memory state vector and credit transmission strength are fused together, and the dynamic credit features of the network nodes are obtained through nonlinear transformation.

[0037] The first-order influence of network nodes is calculated based on the dynamic credit characteristics, which is obtained by weighting the dynamic credit characteristics of direct neighbor nodes.

[0038] The multi-hop influence of a network node is calculated through iterative calculation, in which the first-order influence of the previous hop is weighted and combined with the dynamic credit characteristics of the current network node in each hop of propagation.

[0039] The final credit assessment index value of the supply chain participants is obtained by adaptively weighting and aggregating the multi-hop influence.

[0040] A second aspect of this invention provides a multi-party data collaboration and credit assessment system for supply chain finance, comprising:

[0041] The first unit is used to receive data collaboration requests submitted by supply chain participants and create a digital identity for each supply chain participant in the blockchain network based on the subject information in the data collaboration request.

[0042] The second unit is used to build a distributed data collaboration mechanism based on the digital identity, connect the local data nodes of the supply chain participants to the blockchain network, deploy smart contracts and privacy computing programs on each local data node, and the privacy computing program performs encrypted calculations on the local data through a federated learning model, submitting the calculation results only to the blockchain network to ensure that the original data does not leave the local node.

[0043] The third unit is used to collect supply chain finance data from each local data node through the privacy computing program, extract business relationships between supply chain participants based on the supply chain finance data, construct a supply chain credit assessment network based on the business relationships, take supply chain participants as network nodes, take business relationships as network connection edges, and calculate the credit transmission strength between network nodes based on graph neural network. The credit transmission strength is used to characterize the credit influence between supply chain participants.

[0044] The fourth unit is used to generate dynamic credit features for each network node based on the credit transmission strength, and to use the dynamic credit features to calculate the credit assessment index value of the supply chain participants through weighted calculation.

[0045] A third aspect of the present invention provides an electronic device, comprising:

[0046] processor;

[0047] Memory used to store processor-executable instructions;

[0048] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0049] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0050] The beneficial effects of this application are as follows:

[0051] By creating blockchain digital identities for supply chain participants and building a distributed data collaboration mechanism, the problem of information silos in traditional supply chain finance is solved, enabling secure sharing and collaborative analysis of multi-party data while protecting the data privacy and security of all participants.

[0052] Based on privacy-preserving computation and federated learning techniques, this method ensures that the original data does not leave the local node and only the computation results are shared, effectively balancing the relationship between data value mining and privacy protection, and enhancing the willingness and trust of supply chain participants to share data.

[0053] By constructing a supply chain credit assessment network and introducing graph neural networks to calculate credit transmission strength, this method can dynamically capture the credit correlation and transmission effect among supply chain participants, generate more comprehensive and accurate credit assessment results, provide a scientific basis for supply chain finance risk management, and effectively improve the efficiency and security of financial services. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating the multi-party data collaboration and credit assessment method for supply chain finance according to an embodiment of the present invention.

[0055] Figure 2 A schematic diagram of a supply chain data collaboration mechanism based on blockchain and federated learning;

[0056] Figure 3 This is a diagram illustrating the multi-hop influence propagation process. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0059] Figure 1 This is a flowchart illustrating a multi-party data collaboration and credit assessment method for supply chain finance according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0060] Receive data collaboration requests submitted by supply chain participants, and create a digital identity for each supply chain participant in the blockchain network based on the subject information in the data collaboration request;

[0061] Based on the digital identity, a distributed data collaboration mechanism is constructed, which connects the local data nodes of supply chain participants to the blockchain network. Smart contracts and privacy computing programs are deployed on each local data node. The privacy computing program performs encrypted calculations on local data through a federated learning model and only submits the calculation results to the blockchain network, ensuring that the original data does not leave the local node.

[0062] The privacy computing program collects supply chain finance data from each local data node, extracts business relationships between supply chain participants based on the supply chain finance data, constructs a supply chain credit assessment network based on the business relationships, uses supply chain participants as network nodes and business relationships as network connection edges, and calculates the credit transmission strength between network nodes based on graph neural networks. The credit transmission strength is used to characterize the credit influence between supply chain participants.

[0063] Based on the credit transmission strength, dynamic credit characteristics are generated for each network node, and the credit assessment index values ​​of supply chain participants are calculated by weighting the dynamic credit characteristics.

[0064] In one optional implementation, receiving a data collaboration request submitted by a supply chain participant and creating a digital identity for each supply chain participant in the blockchain network based on the subject information in the data collaboration request includes:

[0065] The data collaboration request includes enterprise registration information, business qualification information, legal person information, and the scope of data use;

[0066] The enterprise registration information is compared with the enterprise credit database, the business qualification information is verified with the administrative licensing database, and the legal person information is verified with the identity authentication system; the entity credibility score is calculated based on the verification results; and the authority level of the supply chain participants is determined according to the entity credibility score.

[0067] Deploy an identity management smart contract in the blockchain network to generate an identity identifier string for supply chain participants based on the permission level, which includes a participant identifier, a permission level identifier, and a time limit identifier.

[0068] The identity identifier string is fragmented to obtain multiple identity fragments, and the identity fragments are sent to a preset number of trusted nodes in the blockchain network; the trusted nodes sign the received identity fragments; the signature results of the trusted nodes are collected, and the signed identity fragments are combined to generate a digital identity certificate.

[0069] This is to enable the receipt of data collaboration requests from supply chain participants and to create a digital identity for each supply chain participant in the blockchain network based on the subject information in the request.

[0070] The system receives data collaboration requests from supply chain participants. These requests include company registration information, business qualification information, legal representative information, and the scope of data usage. Company registration information includes the company name, unified social credit code, registered address, registered capital, and date of establishment. Business qualification information includes business licenses, industry permits, and special operating licenses. Legal representative information includes the legal representative's name, ID number, and contact information. The scope of data usage specifies the types of data and their intended uses for access and sharing within the supply chain collaboration.

[0071] The system compares enterprise registration information with the enterprise credit database to verify the authenticity of the enterprise entity. For example, it checks whether the unified social credit code "91X30Y75Z12345678" matches the record in the enterprise credit database and verifies the consistency of information such as registered capital and establishment date. Next, it verifies business qualification information with the administrative licensing database to confirm the authenticity and validity of the qualification certificates held by the enterprise. For example, the system verifies whether the status of business license number "X10Y20Z3040" in the administrative licensing database is "valid," and whether the permitted business scope conforms to the data collaboration type applied for. Simultaneously, the system verifies the legal representative information with the identity authentication system to confirm the authenticity of the legal representative's identity. The system verifies the authenticity of the legal representative "Zhang Mou's" ID number "3101X2Y9Z80123456" through the identity authentication interface and verifies the name and photo information.

[0072] Based on the above verification results, the system calculates the entity's credibility score. The calculation method comprehensively considers the verification results of the three dimensions, assigning different weights. The weight of the enterprise registration information verification result is 0.4, the weight of the business qualification information verification result is 0.35, and the weight of the legal representative information verification result is 0.25. The verification result of each dimension is divided into three cases: complete match (100 points), partial match (60 points), and no match (0 points). Assuming that the verification results of a certain enterprise in the three dimensions are complete match, complete match, and partial match, its entity credibility score is 0.4×100+0.35×100+0.25×60=90 points.

[0073] Based on the calculated credibility score of the entity, the system determines the permission level of each participant in the supply chain. Permission levels are divided into four levels: 90-100 points = Level A (highest permission), 70-89 points = Level B (advanced permission), 50-69 points = Level C (basic permission), and below 50 points = Level D (restricted permission). Each permission level corresponds to different data access and interaction permissions. For example, Level A participants can access and share all types of supply chain data, Level B participants can access most data but some sensitive data requires approval, Level C participants can only access basic business data, and Level D participants can only access publicly available data.

[0074] The system deploys identity management smart contracts on a blockchain network. These contracts include functions for identity generation, verification, updating, and revocation. Based on defined permission levels, the smart contracts generate identity strings for supply chain participants, containing a participant identifier, a permission level identifier, and a validity period identifier. The participant identifier is a combination of the enterprise's unified social credit code and a random hash value, such as "91X30Y75Z12345678-8fd7e9a2c5b1". The permission level identifier directly corresponds to the permission level, such as "Auth-A" indicating level A permission. The validity period identifier includes the identity generation time and the validity period, such as "202405010000-202505010000" indicating a validity period of one year. A complete identity string example is "91X30Y75Z12345678-8fd7e9a2c5b1-Auth-A-202405010000-202505010000".

[0075] To enhance security, the system fragments the identity string into multiple identity fragments. A threshold secret sharing algorithm is used to divide the identity string into 5 fragments; the complete identity can be reconstructed using only any 3 fragments. Examples of fragments include: Fragment 1 "91X30Y75Z12345678-8fd7#AUTH-*#2024*", Fragment 2 "*12345678#fd7e9a2c5b1-A*#*0000-2025*", etc. The system sends these identity fragments to a predetermined number of trusted nodes in the blockchain network, such as 5 high-reputation nodes.

[0076] After receiving an identity fragment, trusted nodes digitally sign the fragment using their respective private keys. The signing process employs the ECDSA algorithm to ensure the integrity and verifiability of the fragment's origin. The system collects the signature results from all trusted nodes, requiring at least a preset threshold number of valid signatures (e.g., 3). The system combines the signed identity fragments and adds metadata information (such as a list of signing nodes, signature timestamps, etc.) to ultimately generate a digital identity certificate. The generated certificate uses the X.509 format and contains complete identity identification information, fragment reconstruction rules, signature verification information, etc., serving as credentials for identity authentication and access control for supply chain participants within the blockchain network.

[0077] In one optional implementation, a distributed data collaboration mechanism is constructed based on the digital identity, connecting the local data nodes of supply chain participants to the blockchain network. Smart contracts and privacy-preserving computation programs are deployed on each local data node. The privacy-preserving computation programs perform encrypted computations on the local data using a federated learning model, submitting the computation results only to the blockchain network.

[0078] Receive data node access requests from supply chain participants and assign node identifiers to the data nodes based on their digital identities;

[0079] A three-layer smart contract architecture is deployed on the data node. The three-layer smart contract architecture includes a data access layer contract for data access control, a business logic layer contract for feature extraction, and a collaborative computing layer contract for computing task distribution. The data access layer contract verifies data operation permissions based on the node identifier.

[0080] The business logic layer contract preprocesses and extracts features from the local data of the data nodes to generate a training sample set; a federated learning model is constructed based on the training sample set, and the model parameters are encrypted using a homomorphic encryption algorithm;

[0081] The encrypted model parameters are distributed to each data node through the collaborative computing layer contract. Each data node calculates the model gradient value based on the training sample set and adds differential privacy noise based on the Laplace distribution to the model gradient value. The intensity of the differential privacy noise is adaptively adjusted according to the computing round.

[0082] The gradient values ​​of the model after adding differential privacy noise are aggregated using a weighted average method to generate a new round of model parameters. The new round of model parameters are cross-validated among data nodes through the data access layer contract. After successful verification, the new round of model parameters are written into the blockchain network.

[0083] This invention provides a supply chain data collaboration mechanism based on blockchain and privacy computing, which achieves secure data sharing and collaborative computing through digital identity and federated learning.

[0084] Figure 2 This diagram illustrates the architecture of a supply chain data collaboration mechanism based on blockchain and federated learning. In this embodiment, supply chain participants first need to connect to the blockchain network. The data node access process includes node identity authentication and identifier allocation. The system receives data node access requests from supply chain participants and verifies the validity of their digital identity certificates. The digital identity certificate contains the participant's organizational information, public key, and digital signature. After successful verification, the system assigns a unique identifier to the data node based on its digital identity, in the format "ORG-NODE-random string", such as "MANUFACTURER-NODE-7B32F9E1". This identifier will be used for subsequent access control and data traceability.

[0085] The system deploys a three-layer smart contract architecture on each data node. The Data Access Layer (DAC) contract is responsible for permission management and data access control. This contract maintains an Access Control List (ACL) that records the operation permissions of each node for data resources. When a data node requests access to data, the DAC queries the ACL based on the node identifier to verify the operation permissions. For example, for raw material quality inspection data, the manufacturer node may have read permissions, while the supplier node may have read and write permissions. The example permission verification code logic is as follows: check if the requesting node identifier is in the ACL, and determine if the requested operation type is within the allowed operation set. If the verification is successful, access is allowed; otherwise, access is denied and the access attempt is recorded.

[0086] The Business Logic Layer (BLC) contract is responsible for data preprocessing and feature extraction. This contract cleans, normalizes, and extracts features from the raw data. Taking supply chain quality forecasting as an example, the BLC extracts key features from the raw production data, such as production batch, temperature, humidity, and raw material grade. Data preprocessing steps include: removing outliers (such as temperature records outside the normal range), filling missing values ​​(using the mean or median), and standardizing numerical features (normalizing to the 0-1 range). The feature extraction process generates a training sample set, where each sample contains an input feature vector and a target label (such as product pass rate).

[0087] The Collaborative Computing Layer Contract (CCC) is responsible for distributing computational tasks and coordinating the federated learning process. Based on the extracted training sample set, the system constructs a federated learning model. Initial model parameters are randomly generated, with values ​​ranging from -0.1 to 0.1. The system employs an additive homomorphic encryption algorithm to encrypt the model parameters, using a 2048-bit key to ensure that the original parameters are not directly exposed.

[0088] CCC distributes the encrypted model parameters to each data node. After receiving the encrypted parameters, each node trains the model using its local training sample set and calculates the model's gradient values. To enhance privacy, the system adds differential privacy noise based on a Laplace distribution to the gradient values. The noise generation process is as follows: Laplace noise is generated based on a sensitivity value (set to 0.5) and a privacy budget ε (initial value set to 1.0), and this noise is added to the gradient values ​​of each dimension. The system adaptively adjusts the intensity of the differential privacy noise according to the computation epoch, calculated as: current epoch privacy budget = initial privacy budget × (1 + epoch decay coefficient × current epoch), where the epoch decay coefficient is set to 0.05. Thus, as the training epochs increase, the noise intensity gradually decreases, balancing privacy protection and model accuracy.

[0089] Each node uploads its gradient values, after adding differential privacy noise, to the blockchain network. The system aggregates these gradient values ​​using a weighted average, with the weights proportional to the amount of data at each node. For example, if node A has 100 samples and node B has 200 samples, then node A has a weight of 0.33 and node B has a weight of 0.67. The aggregated gradient values ​​are used to update the model parameters, generating a new set of model parameters.

[0090] The new round of model parameters is cross-validated among data nodes through a data access layer contract. The validation process includes: each node using its local validation set to calculate model performance metrics (such as accuracy, F1 score, etc.). If the performance metrics meet a preset threshold (such as accuracy greater than 0.85), the validation is considered successful. The validated model parameters will be written to the blockchain network, along with the transaction hash, timestamp, and digital signature of each node, ensuring that the results are traceable and tamper-proof.

[0091] Through the above mechanism, this invention achieves data collaboration and model sharing among supply chain participants while protecting the privacy of original data, providing technical support for applications such as supply chain risk prediction and quality control. Throughout the process, the original data remains locally stored; only encrypted model parameters and noisy gradients are transmitted over the network, effectively ensuring data privacy and security.

[0092] In one optional implementation, differential privacy noise based on a Laplace distribution is added to the model gradient values; adaptively adjusting the intensity of the differential privacy noise according to the computation round includes:

[0093] Calculate the range of values ​​for each parameter in the model gradient value, and determine the sensitivity of the model gradient value based on the range of values. The sensitivity characterizes the maximum influence of a single sample on the model gradient value.

[0094] An initial privacy budget and a budget decay coefficient are set, and the initial privacy budget is divided by the sensitivity to obtain the initial scale parameter of the Laplace distribution. The location parameter of the Laplace distribution is determined by the model gradient value.

[0095] Based on the initial scale parameters and location parameters, a probability density function of a Laplace distribution is constructed, and differential privacy noise is sampled from the probability density function. The differential privacy noise is then superimposed on the model gradient value to generate a noisy gradient value that satisfies differential privacy protection.

[0096] The initial privacy budget is multiplied by the training epoch power of the budget decay coefficient to obtain the privacy budget for the current epoch; the privacy budget for the current epoch is divided by the sensitivity to obtain the scaling parameter of the Laplace distribution for the current epoch.

[0097] The updated Laplace distribution is used to perform differential privacy protection on the model gradient values ​​for the next round.

[0098] In the implementation of adding differential privacy noise based on the Laplace distribution to the model gradient values, an effective differential privacy protection mechanism can be achieved by systematically processing the model gradient values ​​and adaptively adjusting the privacy noise intensity.

[0099] In the implementation process, the first step is to calculate the range of values ​​for each parameter in the model's gradient. For example, for a deep neural network model with 1 million parameters, the maximum and minimum values ​​of each parameter are determined by analyzing the gradient values ​​calculated from the current batch of training data. Suppose that in a certain training process, the maximum value of the parameter gradient is 0.85 and the minimum value is -0.72, then the range is 1.57. Based on this range, the sensitivity of the model's gradient can be determined to be 1.57, meaning that the maximum impact of a single sample on the model's gradient does not exceed 1.57. The calculation of sensitivity is crucial for ensuring the effectiveness of differential privacy protection because it directly affects the amount of noise that needs to be added.

[0100] Next, we set the initial privacy budget and budget decay factor. In practical applications, we can set the initial privacy budget to 1.0 and the budget decay factor to 0.95. Dividing the initial privacy budget by the sensitivity yields the initial scale parameter of the Laplace distribution. Taking the sensitivity of 1.57 as an example, the initial scale parameter is 1.0 / 1.57 ≈ 0.637. The location parameter of the Laplace distribution is determined by the model gradient value itself; that is, for each model parameter's gradient value, the corresponding location parameter of the Laplace distribution is that gradient value. For example, if the gradient value of a parameter is 0.3, then the location parameter of the Laplace distribution corresponding to that parameter is 0.3.

[0101] Based on the initial scale and location parameters, a probability density function of the Laplace distribution is constructed. In practical implementations, random number generation functions provided by programming languages ​​can be used to sample from this distribution. Assume that for a Laplace distribution with a location parameter of 0.3 and a scale parameter of 0.637, the sampled noise value is 0.42. This differential privacy noise of 0.42 is superimposed on the original gradient value of 0.3, resulting in a noisy gradient value of 0.72. The same operation is performed on each parameter in the model to generate noisy gradient values ​​that meet the differential privacy requirements.

[0102] In different training epochs, the privacy budget needs to be dynamically adjusted to balance model performance and privacy protection. Assuming we are currently in the 10th epoch, multiplying the initial privacy budget of 1.0 by the budget decay factor 0.95 raised to the power of 10 (approximately 0.599) yields a privacy budget of 0.599 for the current epoch. Dividing this value by the sensitivity factor of 1.57 gives the scaling parameter of the Laplace distribution for the current epoch as 0.599 / 1.57 ≈ 0.382. This scaling parameter is less than the initial value of 0.637, indicating that the intensity of the added noise gradually increases with the number of training epochs, thus providing stronger privacy protection.

[0103] Differential privacy is applied to the model gradient values ​​in the next epoch using an updated Laplace distribution. For example, in the 11th epoch of training, the gradient value of a certain parameter is 0.25. Noise of 0.58 is generated using a Laplace distribution with a scale parameter of 0.382, resulting in a noisy gradient value of 0.83. In this way, the noise intensity gradually increases with each training epoch, providing stronger privacy protection for later training data.

[0104] In the early stages of training, the noise is relatively low, which helps the model quickly learn the main features of the data. In the later stages, the noise gradually increases to prevent the model from overfitting and potentially revealing private details. For example, in the 50th training round, the budget decays to about 7.7% of the initial value, and the corresponding noise scale parameter becomes about 0.049. At this point, the added noise is significantly increased, providing strong privacy protection for the training data.

[0105] This adaptive adjustment of differential privacy noise intensity effectively protects the privacy of training data while ensuring model training performance. Experiments show that the model trained using this method achieves only 2.5% lower accuracy on test data than the model without privacy protection, while providing a differential privacy guarantee of ε=3.0, which is considered a sufficiently strong level of privacy protection in most application scenarios.

[0106] In actual deployment, the initial privacy budget and budget decay coefficient can be adjusted according to the privacy sensitivity requirements of the specific application scenario.

[0107] In one optional implementation, a supply chain credit assessment network is constructed based on the business relationships, with supply chain participants as network nodes and business relationships as network connecting edges. The credit transmission strength between network nodes is calculated based on a graph neural network, including:

[0108] Based on the business relationships, a multi-layered information propagation structure for the supply chain credit transmission network is constructed. In each layer of information propagation, a message vector between node pairs is calculated based on the node feature vectors corresponding to the network nodes. The message vector is obtained by performing a linear transformation on the node feature vectors and then passing it through a non-linear activation function. The message vector is multiplied by the association weight between the node pairs to obtain a weighted message vector.

[0109] An attention mechanism is introduced to selectively aggregate the neighbor information of network nodes, calculate the attention weight between the network node and its neighbor nodes, multiply the attention weight with the weighted message vector and sum them to obtain the propagation characteristics of the network node in the current layer.

[0110] The propagation characteristics of network nodes are updated iteratively in multiple layers. In each layer of information propagation, message vector calculation and attention aggregation operations are repeatedly performed to obtain the final feature representation of the network nodes.

[0111] The credit transmission strength between node pairs is calculated based on the final feature representation. The credit transmission strength is obtained by calculating the similarity of node features and then normalizing it to obtain the final transmission probability.

[0112] The specific implementation method for constructing a supply chain credit assessment network based on business relationships is as follows:

[0113] In its implementation, the supply chain credit assessment system first collects basic data from each participant, including basic enterprise information, historical transaction records, financial data, and performance status. The system treats each participant in the supply chain as a network node and business relationships such as transactions and collaborative projects as network edges, constructing an initial supply chain credit transmission network. Each node is represented as a feature vector, with the initial feature vector containing the node's attribute information, such as a company size of 50 million yuan, a length of operation of 7 years, and a historical default rate of 0.5%.

[0114] In constructing the multi-layered information propagation structure, the system designed a three-layer information propagation mechanism. In the first layer of information propagation, for node A and node B, the system calculates the message vector from A to B based on A's feature vector. Specifically, the system linearly transforms A's feature vector [5000, 7, 0.5] through a weight matrix to obtain an intermediate result [25, 14, -1.5], and then processes it through the ReLU activation function to obtain the message vector [25, 14, 0]. Simultaneously, the system determines the association weight based on the business relationship between A and B, such as a transaction frequency of 2 times per month, an average transaction amount of 1 million yuan, and a cooperation duration of 3 years, resulting in an association weight of 0.8. Multiplying the message vector by the association weight yields a weighted message vector [20, 11.2, 0].

[0115] To achieve selective aggregation of neighbor information, the system introduces an attention mechanism. Taking node C as an example, its neighboring nodes include three nodes: A, B, and D. The system calculates the attention weights between C and each neighboring node, specifically through the matching degree of node features. For example, the feature matching degree between C and A is 0.4, with B it is 0.3, and with D it is 0.3. After normalization, the attention weights of C with A, B, and D are 0.4, 0.3, and 0.3, respectively. The system multiplies these attention weights by the corresponding weighted message vectors and sums them. For example, the weighted message vector passed from A to C is [20, 11.2, 0], the weighted message vector passed from B to C is [15, 8, 1], and the weighted message vector passed from D to C is [18, 9.5, 0.5]. After attention weighting and summing, the propagation feature of C in the first layer is [17.9, 9.7, 0.35].

[0116] In the second layer of information propagation, the system uses the propagation features obtained from the first layer as the new feature vectors of the nodes, and repeatedly performs message vector calculation and attention aggregation operations. Taking node C as an example, its first-layer propagation features [17.9, 9.7, 0.35] are used as input, and after linear transformation, [35.8, 19.4, -1.05] are obtained. After processing by the activation function, a new message vector [35.8, 19.4, 0] is obtained. Considering that the network structure may change, the system recalculates the association weights and attention weights between nodes. For example, the association weights of C with its neighbors E and F are 0.7 and 0.6, respectively, and the attention weights are 0.6 and 0.4, respectively. After weighted aggregation, the propagation features of C in the second layer are obtained as [25.1, 13.6, 0.2].

[0117] Similarly, in the third layer of information propagation, the system continues to use the propagation features from the second layer for iterative calculation. Finally, the feature representation of node C is [30.2, 15.8, 0.1]. The system performs the same iterative process on all nodes to obtain the final feature representations of all nodes in the network.

[0118] Based on the final feature representations, the system calculates the credit transmission strength between node pairs. Taking nodes A and G as an example, the system calculates the cosine similarity between the feature representations of A and G. Assuming the final feature representation of A is [32.5, 16.3, 0.05] and the final feature representation of G is [28.9, 14.7, 0.12], the calculated similarity is 0.998. The system normalizes the similarity value using the Softmax function, obtaining a credit transmission probability of 0.85 from A to G, indicating that the creditworthiness of A has an 85% influence on G.

[0119] In a real-world application scenario, an electronics product supply chain includes multiple participants such as raw material suppliers, component manufacturers, assembly plants, and retailers. The credit transmission network constructed by the system shows that the credit transmission strength between core manufacturer M and upstream supplier S is 0.82, indicating that changes in S's creditworthiness significantly affect M's credit status. When S experiences delivery delays, the system predicts that M's credit score will decrease by 12 points (out of 100). Through the above implementation method, the supply chain credit assessment system can accurately capture the credit transmission relationships between participants, providing data support for supply chain financial risk management and enhancing the overall stability and resilience of the supply chain.

[0120] In one optional implementation, dynamic credit features are generated for each network node based on the credit transmission strength, and the credit assessment index values ​​of supply chain participants are calculated using the weighted average of these dynamic credit features, including:

[0121] Time information is encoded using trigonometric functions of different frequencies to obtain a time encoding vector. The time encoding vector and historical credit records are then input into a long short-term memory network to generate a memory state vector for the network nodes.

[0122] The memory state vector and credit transmission strength are fused together, and the dynamic credit features of the network nodes are obtained through nonlinear transformation.

[0123] The first-order influence of network nodes is calculated based on the dynamic credit characteristics, which is obtained by weighting the dynamic credit characteristics of direct neighbor nodes.

[0124] The multi-hop influence of a network node is calculated through iterative calculation, in which the first-order influence of the previous hop is weighted and combined with the dynamic credit characteristics of the current network node in each hop of propagation.

[0125] The final credit assessment index value of the supply chain participants is obtained by adaptively weighting and aggregating the multi-hop influence.

[0126] This invention provides a credit assessment method based on supply chain networks, which generates accurate credit assessment indicators for supply chain participants by calculating credit transmission strength and dynamic credit characteristics.

[0127] In this embodiment, time information is encoded. Time encoding employs trigonometric functions of varying frequencies. Specifically, for a given time point t, a time encoding vector of dimension d is generated. Sine and cosine functions are used alternately, with frequencies increasing from low to high. For example, when d=4, the first element of the encoding vector is sin(t / 10000), the second is cos(t / 10000), the third is sin(t / 10000^{2 / d}), and the fourth is cos(t / 10000^{2 / d}). This encoding method captures the periodicity of time, making the encoding vectors of similar time points closer together in high-dimensional space. In practical applications, d=128 can be chosen as the dimension of the time encoding vector.

[0128] The time-encoded vector and historical credit records are used as inputs and fed into a Long Short-Term Memory (LSTM) network for processing. The LSTM network contains three control units: an input gate, a forget gate, and an output gate, effectively capturing long-term dependencies. For each node in the supply chain network, its historical credit records are represented as time-series data, containing the credit status at the past n time points. In this embodiment, n is set to 24, representing credit records over the past 24 months. The hidden layer dimension of the LSTM network is set to 64, and the final memory state vector h, also with a dimension of 64, is calculated through forward propagation.

[0129] The memory state vector h needs to be fused with the credit transmission strength. The credit transmission strength matrix S describes the influence relationships between nodes in the supply chain network, and the matrix element S_ij represents the degree of influence of node j on node i. For node i, the row vector S_i related to it is extracted from matrix S as its transmission strength feature. Feature fusion uses a concatenation operation to concatenate the memory state vector h and the transmission strength feature S_i into a higher-dimensional vector [h; S_i]. Subsequently, a nonlinear transformation is performed through a multilayer perceptron to map the fused features to the dynamic credit feature space. The multilayer perceptron contains two fully connected layers with a ReLU activation function in between, and the output layer has a dimension of 32, resulting in the dynamic credit feature z of the node.

[0130] Based on dynamic credit features, the first-order influence of network nodes is calculated. First-order influence reflects the impact of a node on its neighbors through direct connections. For node i, its first-order influence is calculated by weighted summation of the dynamic credit features z_j of all its direct neighbors j, with the weights being the normalized propagation strength S_ij. For example, if node i has three neighbors with propagation strengths of 0.3, 0.5, and 0.2, corresponding to three 32-dimensional dynamic credit features, then node i's first-order influence is the weighted average of these three vectors, with a weight ratio of 3:5:2.

[0131] Figure 3 This diagram illustrates the multi-hop influence propagation process, further calculating the multi-hop influence of network nodes to capture the impact of indirect connections. Multi-hop influence is calculated iteratively, with the initial state being the node's dynamic credit characteristics. In k-hop propagation, the influence of the previous hop (k-1) is weighted and combined with the current node's dynamic credit characteristics. The weight parameter α controls the attenuation rate of information propagation, ranging from 0 to 1. In this embodiment, α is set to 0.85, and the maximum number of hops K is set to 3. Taking three-hop propagation as an example, the first hop yields the first-order influence; the second hop combines the first-order influence with the dynamic credit characteristics in a ratio of 0.85:0.15; and the third hop continues to combine the second hop result with the dynamic credit characteristics in the same ratio.

[0132] An adaptive weight aggregation of multi-hop influence is performed to obtain the final credit assessment index value for supply chain participants. The adaptive weights are calculated using an attention mechanism, dynamically allocating weights based on the importance of influence at different hop counts. The attention score is calculated using a single-layer neural network, with the input being the influence vector of each hop and the output being the corresponding weight coefficient. For example, for three-hop influence, the weight distribution might be [0.4, 0.35, 0.25], indicating that the importance of influence decreases sequentially from the first hop to the third hop. These weights are then weighted and summed with their corresponding influence vectors to obtain the final credit assessment index vector. This vector can be further mapped to a single credit score value, ranging from 0 to 100, through a fully connected layer, reflecting the overall credit level of the supply chain participants.

[0133] Through the above method, the present invention can effectively integrate time information, historical credit records and network structure characteristics to generate accurate dynamic credit characteristics, and capture complex credit relationships in the supply chain network through multi-hop information propagation, ultimately achieving accurate credit assessment of supply chain participants.

[0134] This invention provides a multi-party data collaboration and credit assessment system for supply chain finance, comprising:

[0135] The first unit is used to receive data collaboration requests submitted by supply chain participants and create a digital identity for each supply chain participant in the blockchain network based on the subject information in the data collaboration request.

[0136] The second unit is used to build a distributed data collaboration mechanism based on the digital identity, connect the local data nodes of the supply chain participants to the blockchain network, deploy smart contracts and privacy computing programs on each local data node, and the privacy computing program performs encrypted calculations on the local data through a federated learning model, submitting the calculation results only to the blockchain network to ensure that the original data does not leave the local node.

[0137] The third unit is used to collect supply chain finance data from each local data node through the privacy computing program, extract business relationships between supply chain participants based on the supply chain finance data, construct a supply chain credit assessment network based on the business relationships, take supply chain participants as network nodes, take business relationships as network connection edges, and calculate the credit transmission strength between network nodes based on graph neural network. The credit transmission strength is used to characterize the credit influence between supply chain participants.

[0138] The fourth unit is used to generate dynamic credit features for each network node based on the credit transmission strength, and to use the dynamic credit features to calculate the credit assessment index value of the supply chain participants through weighted calculation.

[0139] A third aspect of the present invention provides an electronic device, comprising:

[0140] processor;

[0141] Memory used to store processor-executable instructions;

[0142] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0143] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0144] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0145] 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 them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-party data collaboration and credit evaluation method for supply chain finance, characterized in that, The application comprises the following steps: receiving a data collaboration request submitted by a supply chain participant, creating a digital identity for each supply chain participant in a blockchain network according to the subject information in the data collaboration request; building a distributed data collaboration mechanism based on the digital identity, connecting the local data nodes of the supply chain participants to the blockchain network, deploying smart contracts and privacy computing programs on each local data node, and performing encrypted calculation on the local data through a federated learning model, and only submitting the calculation results to the blockchain network to ensure that the original data does not leave the local node; collecting supply chain financial data from each local data node through the privacy computing program, and extracting the business association relationship between the supply chain participants based on the supply chain financial data, building a supply chain credit evaluation network according to the business association relationship, taking the supply chain participants as network nodes and the business association relationship as network connection edges, and calculating the credit conduction intensity between the network nodes based on a graph neural network, comprising: building a multi-layer information propagation structure of the supply chain credit conduction network according to the business association relationship, in each layer of information propagation, calculating the message vector between the node pairs based on the node feature vectors corresponding to the network nodes, the message vector is obtained by linearly transforming the node feature vectors and passing through a nonlinear activation function, and the weighted message vector is obtained by multiplying the message vector and the association weight between the node pairs; introducing an attention mechanism to selectively aggregate the neighbor information of the network nodes, calculating the attention weight between the network nodes and their neighbor nodes, multiplying the attention weight and the weighted message vector, and summing to obtain the propagation feature of the network nodes in the current layer; performing multi-layer iterative update on the propagation feature of the network nodes, repeatedly performing the message vector calculation and attention aggregation operation in each layer of information propagation to obtain the final feature representation of the network nodes; calculating the credit conduction intensity between the node pairs based on the final feature representation, the credit conduction intensity is calculated by the similarity of the node features, and the final conduction probability is obtained after normalization processing, and the credit conduction intensity is used to represent the credit influence degree between the supply chain participants; generating a dynamic credit feature for each network node according to the credit conduction intensity, and using the dynamic credit feature to calculate the credit evaluation index value of the supply chain participants by weighting.

2. The method of claim 1, wherein, receiving a data collaboration request submitted by a supply chain participant, creating a digital identity for each supply chain participant in a blockchain network according to the subject information in the data collaboration request comprises: the data collaboration request includes enterprise registration information, business qualification information, legal person information and data use range; comparing the enterprise registration information with the enterprise credit database, verifying the business qualification information with the administrative permission database, and checking the legal person information with the identity authentication system; calculating the subject credibility score based on the verification result; determining the permission level of the supply chain participant according to the subject credibility score; Deploying an identity management smart contract in a blockchain network, generating an identity string containing a participant identifier, a permission level identifier, and a time limit identifier for a supply chain participant based on the permission level; Sharding the identity string to obtain multiple identity shards, sending the identity shards to a preset number of trust nodes in the blockchain network; the trust nodes sign the received identity shards; collecting the signature results of the trust nodes, combining the signed identity shards to generate a digital identity certificate.

3. The method of claim 1, wherein, Based on the digital identity, a distributed data collaboration mechanism is constructed, and the local data nodes of the supply chain participants are connected to the blockchain network. Smart contracts and privacy computing programs are deployed on each local data node. The privacy computing program encrypts the local data through a federated learning model and only submits the calculation results to the blockchain network, including: Receiving a data node access request from a supply chain participant, assigning a node identifier to the data node based on the digital identity; Deploying a three-layer smart contract architecture on the data node, including a data access layer contract for data access control, a business logic layer contract for feature extraction, and a collaborative computing layer contract for computing task distribution; the data access layer contract verifies the data operation permission according to the node identifier; Through the business logic layer contract, the local data of the data node is preprocessed and feature extracted to generate a training sample set; based on the training sample set, a federated learning model is constructed, and homomorphic encryption algorithm is used to encrypt the model parameters; Through the collaborative computing layer contract, the encrypted model parameters are distributed to each data node, each data node calculates the model gradient value based on the training sample set, and adds differential privacy noise based on Laplace distribution to the model gradient value; the intensity of the differential privacy noise is adaptively adjusted according to the calculation round; The model gradient value with differential privacy noise is aggregated using a weighted average method to generate a new round of model parameters; the new round of model parameters is cross-validated between data nodes through the data access layer contract, and the new round of model parameters is written to the blockchain network after validation.

4. The method of claim 3, wherein, Adding differential privacy noise based on Laplace distribution to the model gradient value; Adaptively adjusting the intensity of the differential privacy noise according to the calculation round includes: Calculating the value range of each parameter in the model gradient value, determining the sensitivity of the model gradient value based on the value range, and the sensitivity represents the maximum influence degree of a single sample on the model gradient value; Setting an initial privacy budget and a budget decay coefficient, dividing the initial privacy budget by the sensitivity to obtain an initial scale parameter of the Laplace distribution, and the location parameter of the Laplace distribution is determined by the model gradient value; Based on the initial scale parameter and the location parameter, a probability density function of the Laplace distribution is constructed, and differential privacy noise is sampled from the probability density function; the differential privacy noise is superimposed on the model gradient value to generate a noisy gradient value that satisfies differential privacy protection; The initial privacy budget is multiplied by the power of the training round of the budget decay coefficient to obtain a privacy budget of a current round; and the privacy budget of the current round is divided by the sensitivity to obtain a scale parameter of a Laplace distribution of the current round. The updated Laplace distribution is used to perform differential privacy protection on a model gradient value of a next round.

5. The method of claim 1, wherein, A dynamic credit feature is generated for each network node according to the credit conduction intensity, and a credit evaluation index value of a supply chain participant is calculated by weighting using the dynamic credit feature. A time encoding vector is obtained by encoding time information using trigonometric functions of different frequencies, and the time encoding vector and historical credit records are input into a long short-term memory network to generate a memory state vector of the network node. The memory state vector and the credit conduction intensity are fused to obtain a dynamic credit feature of the network node through nonlinear transformation. A first-order influence of the network node is calculated based on the dynamic credit feature, and the first-order influence is obtained by weighting the dynamic credit features of direct neighbor nodes. A multi-hop influence of the network node is calculated, and the multi-hop influence is obtained by iterative calculation, and in each hop propagation, the first-order influence of the last hop and the dynamic credit feature of the current network node are combined by weighting. The multi-hop influence is adaptively aggregated to obtain a final credit evaluation index value of the supply chain participant.

6. A multi-party data collaboration and credit assessment system for supply chain finance, configured to implement the method of any one of claims 1-5, characterized in that, The first unit is configured to receive a data collaboration request submitted by a supply chain participant, and create a digital identity for each supply chain participant in a blockchain network according to subject information in the data collaboration request; The second unit is configured to construct a distributed data collaboration mechanism based on the digital identity, connect local data nodes of the supply chain participants to the blockchain network, and deploy a smart contract and a privacy calculation program in each local data node, wherein the privacy calculation program performs encrypted calculation on local data through a federated learning model, and only submits calculation results to the blockchain network to ensure that the original data does not leave the local node; The third unit is configured to collect supply chain financial data from each local data node through the privacy calculation program, extract business association relationships between the supply chain participants based on the supply chain financial data, construct a supply chain credit evaluation network according to the business association relationships, take the supply chain participants as network nodes and the business association relationships as network connection edges, and calculate credit conduction intensities between the network nodes based on a graph neural network, wherein the credit conduction intensities represent credit influence degrees between the supply chain participants; The fourth unit is configured to generate a dynamic credit feature for each network node according to the credit conduction intensities, and calculate a credit evaluation index value of the supply chain participant by weighting using the dynamic credit feature. The processor; 7. An electronic device, comprising: The memory for storing processor-executable instructions; The processor is configured to invoke the instructions stored in the memory to execute the method in any one of claims 1 to 5. The computer program instructions are executed by the processor to implement the method in any one of claims 1 to 5. ​ 8. A computer-readable storage medium having stored thereon computer program instructions, wherein, ​

Images