Financial science and technology multi-dimensional trust evaluation method and system based on block chain

Abstract

The invention relates to a financial science and technology multi-dimensional trust evaluation method and system based on a block chain, and the method comprises the steps: collecting on-chain transaction data of a financial subject and encrypted and verified off-chain multi-source data through a smart contract disposed in a block chain network; constructing a credit model comprising at least four orthogonal evaluation dimensions, wherein the dimensions comprise on-chain behavior credibility, off-chain asset verification degree, social network influence and historical performance volatility; dynamically adjusting the weight coefficient of each dimension by adopting a time sequence prediction model, and writing a weight adjustment result into the block chain after the weight adjustment result is verified by a consensus mechanism; and generating a final trust evaluation result containing the Merkle proof, and storing a result hash value to a block chain non-tampering database. According to the financial science and technology multi-dimensional trust evaluation method and system based on the block chain, by fusing on-chain behaviors, off-chain assets, social networks and historical performance four-dimensional orthogonal data, the default prediction accuracy is improved, and the long-tail user coverage rate is improved.

Description

Technical Field

[0001] The present invention relates to the field of financial technology trust evaluation technology, and specifically to a financial technology multi-dimensional trust evaluation method and system based on blockchain. Background Art

[0002] With the rapid development of FinTech, financial services are becoming increasingly digital and intelligent, with an increasing number of financial transactions and services being conducted in a digital environment. However, the trustworthiness of FinTech service providers has become a key factor in user selection.

[0003] For example, Chinese patent 202410093276.2 proposes a blockchain-based multi-dimensional trust evaluation method for financial technology. This method utilizes blockchain technology and a trust evaluation weighting algorithm to store and weight financial transaction trust assessment results to generate a multi-dimensional trust evaluation result. This method then performs detection and early warning on these results to implement corresponding risk control evaluation decisions. This method can accurately reflect the true credibility of financial technology service providers.

[0004] However, existing multi-dimensional trust evaluation methods and systems in FinTech have the following three major flaws:

[0005] 1) Traditional credit reporting relies on centralized institutional data, such as bank statements, which results in data silos.

[0006] 2) Existing blockchain credit evaluations are mostly limited to a single dimension of transaction history;

[0007] 3) Lack of real-time dynamic update mechanism, unable to reflect changes in subject behavior.

[0008] Therefore, a multi-dimensional trust evaluation method and system of financial technology based on blockchain is proposed to solve the above problems. Summary of the Invention

[0009] In response to the shortcomings of the existing technology, the present invention provides a blockchain-based financial technology multi-dimensional trust evaluation method and system, which has the advantages of improving evaluation accuracy, increasing computing efficiency and enhancing privacy protection. It solves the problem that traditional credit investigation relies on centralized institutional data, such as bank statements, and has data silos. The existing blockchain credit evaluation is mostly limited to a single dimension of transaction history, lacks a real-time dynamic update mechanism, and cannot reflect changes in subject behavior.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a blockchain-based multi-dimensional trust evaluation method for financial technology, comprising the following steps:

[0011] S1. Through smart contracts deployed on the blockchain network, collect financial entities' on-chain transaction data and cryptographically verified off-chain multi-source data;

[0012] S2. Construct a credit model that includes at least four orthogonal evaluation dimensions, including: on-chain behavior credibility, off-chain asset verification, social network influence, and historical performance volatility;

[0013] S3. Use the time series prediction model to dynamically adjust the weight coefficients of each dimension. The weight adjustment results are verified by the consensus mechanism and then written into the blockchain.

[0014] S4. Generate the final trust evaluation result including the Merkle proof and store the result hash value in the blockchain immutable database.

[0015] Furthermore, the calculation of the off-chain asset verification degree includes the following steps:

[0016] 1) Verify the authenticity of offline assets through zero-knowledge proof technology without revealing specific values;

[0017] 2) Using an oracle network to access authoritative third-party data sources processed by the TEE trusted execution environment;

[0018] 3) The asset verification results are cross-verified with the on-chain evidence data through a two-factor authentication mechanism.

[0019] Furthermore, the dynamic adjustment of the weight coefficients of each dimension specifically includes the following steps:

[0020] 1) Extract the historical rating matrix of the last N time periods as the input of the LSTM neural network;

[0021] 2) The output layer uses the softmax function to generate new weight distributions for each dimension;

[0022] 3) When a sudden financial event is detected, an emergency weight reset mechanism based on reinforcement learning is triggered;

[0023] The LSTM neural network training process includes:

[0024] Use a loss function with a time decay factor: \mathcal{L}=\sum_{t=1}^T\lambda^{Tt}\|y_t-\hat{y}_t\|^2

[0025] Where λ∈(0,1) is the attenuation coefficient and T is the time window size;

[0026] The training data must include credit evaluation records of at least three typical financial scenarios, and a federated learning mechanism must be used to protect data privacy when updating the model.

[0027] Furthermore, it also includes:

[0028] S5. Select a differentiated combination of evaluation dimensions based on the type of financial entity. For corporate entities, add the dimension of supply chain relevance, and for individual entities, add the dimension of consumer behavior stability.

[0029] S6. Whenever the evaluation result reaches a preset threshold, the financial business strategy executed by the smart contract is automatically triggered.

[0030] Furthermore, the calculation of the social network influence dimension includes:

[0031] 1) Construct a financial entity association graph, where nodes represent entities and edges represent capital transactions, guarantee relationships, or equity relationships;

[0032] 2) An improved PageRank algorithm is used to calculate the influence score, where edge weights are dynamically adjusted based on transaction frequency and amount, and a decay factor is introduced to handle historical correlations.

[0033] The historical contract volatility dimension is quantified as follows:

[0034] 1) Calculate the standard deviation of the fulfillment rate within the sliding time window;

[0035] 2) Weighted treatment of abnormal performance events, where:

[0036] 3) The weight coefficient of delayed performance is α;

[0037] 4) The weight coefficient of the default event is β;

[0038] Among them, β>α>1.

[0039] A blockchain-based multi-dimensional trust evaluation system for financial technology, including:

[0040] Data collection module: includes data crawler components deployed on alliance chain nodes and on-chain data listeners;

[0041] Credit calculation engine: consists of pluggable dimension algorithm modules and dynamic weight adjusters;

[0042] Blockchain evidence storage layer: adopts a main chain-side chain dual structure, the main chain stores the evaluation result summary, and the side chain stores the original data fingerprint;

[0043] Smart contract components: include three types of contracts: automatic execution contracts, data verification contracts, and weight update contracts.

[0044] Furthermore, the blockchain evidence storage layer adopts an improved consensus mechanism, the probability of selecting the accounting node is positively correlated with its historical credit score, and a credit decay factor is introduced to reduce the calculation weight of historical behavior data that has exceeded its validity period. The new data block must contain the cross-validation hash value of the evaluation result of the previous block.

[0045] Furthermore, the credit calculation engine further includes:

[0046] Anomaly detection submodule: Identifies data injection attacks that attempt to manipulate ratings using the isolation forest algorithm;

[0047] Explanatory submodule: Generates a visual scoring basis report that meets regulatory requirements.

[0048] Furthermore, the data collection module also includes an off-chain data verification unit, which is used to verify API interface data using digital signatures, sensor fingerprint authentication for IoT device data, and blockchain evidence + hash comparison dual verification for manually uploaded files;

[0049] The credit calculation engine also includes a cross-chain verification unit, which is used to aggregate multi-chain credit data through a cross-chain protocol when the subject involves multiple blockchain networks, and use threshold signature technology to verify the authenticity of the cross-chain data;

[0050] The blockchain evidence layer also includes a credit shard storage unit, which stores the evaluation data in the public chain according to public data (final score), semi-public data (dimension score) in the alliance chain, and private data (original certificate) in the private chain with authorized access.

[0051] Compared with the existing technology, the technical solution of this application has the following beneficial effects:

[0052] 1. This blockchain-based fintech multi-dimensional trust evaluation method and system improves the accuracy of default prediction and increases the coverage of long-tail users by integrating four-dimensional orthogonal data: on-chain behavior, off-chain assets, social networks, and historical performance.

[0053] 2. This blockchain-based fintech multi-dimensional trust evaluation method and system uses a hybrid algorithm of LSTM+reinforcement learning to reduce the response speed to field fluctuations, which can provide early warning of liquidity risks. It adopts a dual-chain storage architecture, with the main chain ensuring immutability and the side chain reducing storage overhead. By improving the PBFT consensus: nodes with high credit scores are given priority in accounting, which improves consensus efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flowchart of a multi-dimensional trust evaluation method for financial technology based on blockchain in the present invention;

[0055] Figure 2 This is a framework diagram of a multi-dimensional trust evaluation system for financial technology based on blockchain in the present invention. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] Example 1

[0058] See also Figure 1 In this embodiment, a multi-dimensional trust evaluation method for financial technology based on blockchain includes the following steps:

[0059] S1. Through smart contracts deployed on the blockchain network, the on-chain transaction data of financial entities and cryptographically verified off-chain multi-source data are collected. The calculation of the off-chain asset verification degree includes the following steps:

[0060] 1) Verify the authenticity of offline assets through zero-knowledge proof technology without revealing specific values;

[0061] 2) Using an oracle network to access authoritative third-party data sources processed by the TEE trusted execution environment;

[0062] 3) The asset verification results are cross-verified with the on-chain evidence data through a two-factor authentication mechanism;

[0063] S2. Build a credit model that includes at least four orthogonal evaluation dimensions: on-chain behavior credibility, off-chain asset verification, social network influence, and historical performance volatility;

[0064] S3. Use the time series prediction model to dynamically adjust the weight coefficients of each dimension. The weight adjustment results are written into the blockchain after verification through the consensus mechanism. Dynamically adjusting the weight coefficients of each dimension specifically includes the following steps:

[0065] 1) Extract the historical rating matrix of the last N time periods as the input of the LSTM neural network;

[0066] 2) The output layer uses the softmax function to generate new weight distributions for each dimension;

[0067] 3) When a sudden financial event is detected, an emergency weight reset mechanism based on reinforcement learning is triggered;

[0068] S4. Generate the final trust evaluation result including the Merkle proof and store the resulting hash value in the blockchain immutable database;

[0069] S5. Select a differentiated combination of evaluation dimensions based on the type of financial entity. For corporate entities, add the dimension of supply chain relevance, and for individual entities, add the dimension of consumer behavior stability.

[0070] S6. Whenever the evaluation result reaches a preset threshold, the financial business strategy executed by the smart contract is automatically triggered.

[0071] Specifically, the LSTM neural network training process includes:

[0072] Use a loss function with a time decay factor: \mathcal{L}=\sum_{t=1}^T\lambda^{Tt}\|y_t-\hat{y}_t\|^2

[0073] Where λ∈(0,1) is the attenuation coefficient and T is the time window size;

[0074] The training data must include credit evaluation records of at least three typical financial scenarios, and a federated learning mechanism must be used to protect data privacy when updating the model.

[0075] In this embodiment, the hybrid algorithm of LSTM and reinforcement learning is as follows:

[0076] #Dynamic weight adjustment example (actual deployment code snippet)

[0077] def update_weights(scores):

[0078] lstm_output = lstm_model.predict(scores[-12:]) #12 period historical data

[0079] if emergency_event_detected():#Emergency financial event monitoring

[0080] return reinforcement_learning_adjust(lstm_output)

[0081] return softmax(lstm_output).

[0082] In this embodiment, the social network influence dimension is as follows:

[0083] 1) Construct a financial entity association graph, where nodes represent entities and edges represent capital transactions, guarantee relationships, or equity relationships;

[0084] 2) An improved PageRank algorithm is used to calculate the influence score, where edge weights are dynamically adjusted based on transaction frequency and amount, and a decay factor is introduced to handle historical correlations.

[0085] The historical strike volatility dimension is quantified as follows:

[0086] 1) Calculate the standard deviation of the fulfillment rate within the sliding time window;

[0087] 2) Weighted treatment of abnormal performance events, where:

[0088] 3) The weight coefficient of delayed performance is α;

[0089] 4) The weight coefficient of the default event is β;

[0090] Among them, β>α>1.

[0091] Example 2

[0092] Please refer to Figure 2 In this embodiment, a blockchain-based multi-dimensional trust evaluation system for financial technology includes:

[0093] Data collection module: includes data crawler components deployed on alliance chain nodes and on-chain data listeners;

[0094] Credit calculation engine: Consists of pluggable dimension algorithm modules and a dynamic weight adjuster. It includes an anomaly detection submodule that uses the isolation forest algorithm to identify data injection attacks that attempt to manipulate scores, and an interpretive submodule that generates visual scoring reports that meet regulatory requirements.

[0095] Blockchain evidence storage layer: adopts a main chain and side chain dual structure. The main chain stores the evaluation result summary, and the side chain stores the original data fingerprint. It uses an improved consensus mechanism. The probability of selecting the accounting node is positively correlated with its historical credit score. A credit decay factor is introduced to reduce the calculation weight of historical behavior data that has exceeded its validity period. New data blocks must contain the cross-validation hash value of the previous block's evaluation results.

[0096] Smart contract components: include three types of contracts: automatic execution contracts, data verification contracts, and weight update contracts.

[0097] Specifically, the data collection module also includes an off-chain data verification unit, which is used to verify API interface data using digital signatures, sensor fingerprint authentication for IoT device data, and blockchain evidence + hash comparison for manually uploaded files.

[0098] The credit calculation engine also includes a cross-chain verification unit, which is used to aggregate multi-chain credit data through a cross-chain protocol when the subject involves multiple blockchain networks, and use threshold signature technology to verify the authenticity of cross-chain data;

[0099] The blockchain evidence layer also includes a credit shard storage unit, which stores evaluation data in the public chain as public data (final score), semi-public data (dimension score) in the alliance chain, and private data (original certificate) in the private chain with authorized access.

[0100] Specifically, the present invention is compared with the traditional technology as shown in the following table:

[0101]

[0102]

[0103] In this embodiment, the dynamic weight adjustment mechanism adopts a time-decay LSTM model, as follows:

[0104] #Weight dynamic adjustment core algorithm

[0105] def temporal_attention(inputs):

[0106] time_weights = tf.exp(-0.1*tf.range(0,12))#time attenuation coefficient

[0107] weighted_inputs=inputs*tf.reshape(time_weights,(1,12,1))

[0108] lstm_out=LSTM(units=4)(weighted_inputs)

[0109] return Softmax()(lstm_out).

[0110] In summary, this blockchain-based financial technology multi-dimensional trust evaluation method and system, by integrating four-dimensional orthogonal data of on-chain behavior, off-chain assets, social networks, and historical performance, improves the accuracy of default prediction and the coverage of long-tail users. It adopts a hybrid algorithm of LSTM+reinforcement learning to reduce the response speed of field fluctuations, which can provide early warning of liquidity risks. It adopts a dual-chain storage architecture, with the main chain ensuring immutability and the side chain reducing storage overhead. By improving the PBFT consensus: nodes with high credit scores are given priority in accounting, the consensus efficiency is improved.

[0111] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0112] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional trust evaluation method for financial technology based on blockchain, characterized by: The following steps are involved: S1. Through smart contracts deployed on the blockchain network, collect financial entities' on-chain transaction data and cryptographically verified off-chain multi-source data; S2. Construct a credit model that includes at least four orthogonal evaluation dimensions, including: on-chain behavior credibility, off-chain asset verification, social network influence, and historical performance volatility; S3. Use the time series prediction model to dynamically adjust the weight coefficients of each dimension. The weight adjustment results are verified by the consensus mechanism and then written into the blockchain. S4. Generate the final trust evaluation result including the Merkle proof and store the result hash value in the blockchain immutable database.

2. A blockchain-based multi-dimensional trust evaluation method for financial technology according to claim 1, characterized in that: The calculation of the off-chain asset verification degree includes the following steps: 1) Verify the authenticity of offline assets through zero-knowledge proof technology without revealing specific values; 2) Using an oracle network to access authoritative third-party data sources processed by the TEE trusted execution environment; 3) The asset verification results are cross-verified with the on-chain evidence data through a two-factor authentication mechanism.

3. The multi-dimensional trust evaluation method for financial technology based on blockchain according to claim 1 is characterized in that: The dynamic adjustment of the weight coefficients of each dimension specifically includes the following steps: 1) Extract the historical rating matrix of the last N time periods as the input of the LSTM neural network; 2) The output layer uses the softmax function to generate new weight distributions for each dimension; 3) When a sudden financial event is detected, an emergency weight reset mechanism based on reinforcement learning is triggered; The LSTM neural network training process includes: Use a loss function with a time decay factor: \mathcal{L}=\sum_{t=1}^T\lambda^{Tt}\|y_t-\hat{y}_t\|^2 Where λ∈(0,1) is the attenuation coefficient and T is the time window size; The training data must include credit evaluation records of at least three typical financial scenarios, and a federated learning mechanism must be used to protect data privacy when updating the model.

4. The multi-dimensional trust evaluation method for financial technology based on blockchain according to claim 1 is characterized in that: Also includes: S5. Select a differentiated combination of evaluation dimensions based on the type of financial entity. For corporate entities, add the dimension of supply chain relevance, and for individual entities, add the dimension of consumer behavior stability. S6. Whenever the evaluation result reaches a preset threshold, the financial business strategy executed by the smart contract is automatically triggered.

5. The multi-dimensional trust evaluation method for financial technology based on blockchain according to claim 1 is characterized in that: The calculation of the social network influence dimension includes: 1) Construct a financial entity association graph, where nodes represent entities and edges represent capital transactions, guarantee relationships, or equity relationships; 2) An improved PageRank algorithm is used to calculate the influence score, where edge weights are dynamically adjusted based on transaction frequency and amount, and a decay factor is introduced to handle historical correlations. The historical strike volatility dimension is quantified as follows: 1) Calculate the standard deviation of the fulfillment rate within the sliding time window; 2) Weighted treatment of abnormal performance events, where: 3) The weight coefficient of delayed performance is α; 4) The weight coefficient of the default event is β; Among them, β>α>1.

6. A blockchain-based multi-dimensional trust evaluation system for financial technology, comprising a blockchain multi-dimensional trust evaluation system that implements the method described in claims 1-5, characterized in that: The blockchain multi-dimensional trust evaluation system is specifically as follows: Data collection module: includes data crawler components deployed on alliance chain nodes and on-chain data listeners; Credit calculation engine: consists of pluggable dimension algorithm modules and dynamic weight adjusters; Blockchain evidence storage layer: adopts a main chain-side chain dual structure, the main chain stores the evaluation result summary, and the side chain stores the original data fingerprint; Smart contract components: include three types of contracts: automatic execution contracts, data verification contracts, and weight update contracts.

7. A blockchain-based multi-dimensional trust evaluation system for financial technology according to claim 6, characterized in that: The blockchain evidence storage layer adopts an improved consensus mechanism. The probability of selecting an accounting node is positively correlated with its historical credit score. A credit decay factor is introduced to reduce the calculation weight of historical behavior data that has exceeded its validity period. The new data block must contain a cross-validation hash value of the evaluation result of the previous block.

8. The blockchain-based multi-dimensional trust evaluation system for financial technology according to claim 6, characterized in that: The credit calculation engine further includes: Anomaly detection submodule: Identifies data injection attacks that attempt to manipulate ratings using the isolation forest algorithm; Explanatory submodule: Generates a visual scoring basis report that meets regulatory requirements.

9. The blockchain-based multi-dimensional trust evaluation system for financial technology according to claim 6, characterized in that: The data collection module also includes an off-chain data verification unit for digital signature verification of API interface data, sensor fingerprint authentication of IoT device data, and blockchain evidence + hash comparison dual verification of manually uploaded files; The credit calculation engine also includes a cross-chain verification unit, which is used to aggregate multi-chain credit data through a cross-chain protocol when the subject involves multiple blockchain networks, and use threshold signature technology to verify the authenticity of the cross-chain data; The blockchain evidence layer also includes a credit shard storage unit, which stores the evaluation data in the public chain according to public data (final score), semi-public data (dimension score) in the alliance chain, and private data (original certificate) in the private chain with authorized access.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images