Data element credible circulation system based on block chain and artificial intelligence

Abstract

The invention discloses a data element trusted circulation system based on a block chain and artificial intelligence, and relates to the technical field of data circulation. The demand analysis module is used for obtaining data demands of a demand side for feature analysis and feature extraction to obtain demand feature items, and the demand feature items are used for representing the data demands of the demand side based on the trusted space; and the demand retrieval module is used for performing feature retrieval based on the demand feature items to retrieve at least one piece of retrieval data. According to the method, the demand retrieval module, the feature sorting module and the data preheating module are cooperatively implemented, the on-chain operation burden and the real-time query response time are effectively reduced, the demand analysis module adopts a federated learning framework and zero-knowledge proof, intelligent agent demand vector generation and gradient parameter exchange are only performed under the chain, and the real-time query response time is effectively shortened. The demand for directly processing big data on a chain is reduced, and the demand retrieval module sets a retrieval threshold value and splits large-range query into single retrieval.

Description

Technical Field

[0001] This invention relates to the field of data circulation technology, specifically to a trusted data element circulation system based on blockchain and artificial intelligence. Background Technology

[0002] A trusted data element circulation system is a key infrastructure supporting the market-based allocation of data elements. Its core objective is to achieve efficient, smooth, and orderly flow and value release of data elements while ensuring data security, privacy protection, clear ownership, and compliance control. By integrating technology, rules, and mechanisms, it builds a trustworthy environment to solve the problem of trust deficiency in data circulation. Data elements differ from traditional factors of production; their circulation faces security and privacy risks. Data is easy to copy, leak, and abuse due to its low cost, and involves personal privacy and trade secrets. Furthermore, ownership is often ambiguous, and the relationships between data ownership, usage rights, and revenue rights are complex. The trusted circulation system aims to systematically solve these problems and establish a "game rule" and "operating environment" based on mutual trust among all parties.

[0003] A processing method and system for trusted data element circulation technology, with patent publication number CN120147012A, calculates the similarity between a circulating technology text dataset and a trusted technology text dataset, selects trusted circulation technology texts, matches them with data element circulation links, and constructs a trusted circulation technology system that runs through the entire process of data element circulation. It integrates the technologies of each link to form a complete technology system framework, overcomes the problem of existing technologies that only focus on the local and lack overall planning, and enables data elements to receive effective technical support throughout the circulation process.

[0004] In the trusted data element circulation system, as described above and similar technical solutions, when a demander needs to obtain the required data within the trusted space, the system needs to perform feature analysis and retrieval based on the required data, and select the target data that best matches the demander's needs from the numerous retrieval results. However, existing on-chain queries are costly and have high latency, which affects the demander's experience and usability. When the demander's required data volume is large, the query cost and query latency will be further aggravated. Summary of the Invention

[0005] The purpose of this invention is to provide a trusted data element circulation system based on blockchain and artificial intelligence to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a trusted data element circulation system based on blockchain and artificial intelligence, comprising: Demand Analysis Module: Acquires data requirements from the demand side, performs feature analysis and feature extraction to obtain demand feature items, which are used to represent the demand side's data requirements based on the trust space; Demand retrieval module: Based on demand feature items, feature retrieval is performed to retrieve at least one data item and obtain the target dataset. The target dataset is used to represent the associated data in the trust space that matches the features of the demand feature items. Feature ranking module: Based on the demand feature items, the target dataset is ranked to obtain ranking information items. The search data with the first position in the ranking result is used as the benchmark demand data of the demand feature items to obtain the demand feature data that matches the demand feature items. Data preheating module: Based on the feature ranking criteria used to rank the target dataset according to the required feature items, the retrieved data in the auxiliary connection module is sorted a second time. The basis for the second sorting is the weight ratio of the feature ranking criteria used for feature ranking. The data is sorted according to the feature ranking criteria as the main weight, and the results of the second sorting are stored to obtain the stored preheated data, thereby realizing the data preheating and storage of the data combination. Preheating extraction module: The data requirements are obtained again, and feature analysis and feature extraction are performed again to obtain updated feature items. Based on the updated feature items, the module compares them with the auxiliary connection module to obtain the target combination. Based on the target combination, the module maps to the target auxiliary storage page and performs data extraction and presentation, thereby realizing the preheating retrieval and extraction of data.

[0007] Furthermore, the method for obtaining the demand feature items includes: Based on the Trust Space and with blockchain as the foundation of trust technology, the demand side generates a demand vector through an off-chain smart agent and uploads the commitment value verified by zero-knowledge proof to the blockchain to obtain the uploaded information item. Trusted Space uses artificial intelligence to implement a federated learning framework in conjunction with the data platform and the demand side. Based on the uploaded information items, gradient parameters are exchanged and features are extracted in the local model training to obtain the required feature items.

[0008] Furthermore, the method for obtaining the demand feature items also includes: By dynamically selecting participating nodes through a verifiable random function, participating node items are obtained, the demand vector of participating node items is acquired, and the commitment value verified by zero-knowledge proof is uploaded to the blockchain to obtain the uploaded information participating item. The federated learning framework is executed based on the uploaded information participants. Based on the local information of the uploaded information participants, at least two local domains are obtained. Based on artificial intelligence, gradient parameters are exchanged and features are extracted in the local model training through the local domains to obtain the required feature terms.

[0009] Furthermore, the method for obtaining the target dataset includes: Obtain feature type information of the required feature items and feature requirement information corresponding to the feature type information, and perform a single search based on the individual feature type information and the corresponding feature requirement information; Set a retrieval threshold, which is a fixed retrieval difference between the single feature type information and the corresponding feature requirement information, obtain the result set of a single retrieval, and then obtain the target dataset.

[0010] Furthermore, the method for obtaining the sorting information item includes: Based on the demand feature items, feature classification is performed to obtain at least one demand feature type information. The demand restriction threshold corresponding to the demand feature type information is obtained, including quantity information and time information, and the demand feature attribute information corresponding to the demand feature type item is obtained. The target dataset is split based on the demand feature attribute information. Based on the splitting result, the data information in the target dataset that meets the demand feature attribute information is obtained. The matching items of the data information are used as the criterion to sort the features and obtain the sorting result information. Set a rejection threshold, which is a fixed percentage value. Based on the rejection threshold, the last item in the sorting result information is removed by a percentage to obtain the remaining sorting information and thus obtain the sorting information item.

[0011] Furthermore, the secondary sorting method includes: The demand feature data corresponding to the first search data in the sorting information item is obtained as the main judgment feature. Based on the feature splitting results of the target dataset, at least one feature to be judged is obtained. The target dataset is sorted sequentially using the features to be determined as the main determination features, resulting in at least one secondary sorting result.

[0012] Furthermore, the method for acquiring the stored preheating data includes: Based on the combined results of the target dataset, the trusted space creates an auxiliary connection module, which includes at least one auxiliary storage page; Based on the secondary sorting result, the sorting result is independently split according to the main weight category, resulting in at least two secondary sorting items. The secondary sorting items are stored in the secondary storage pages respectively, and the main weight category is used as the secondary tag of the secondary storage page.

[0013] Furthermore, the method for obtaining the target combination includes: Based on the updated feature items, obtain the updated feature type information and the updated feature requirement information corresponding to the updated feature type information. Obtain the update requirement limit threshold of the updated feature requirement information, including quantity information and time information. Compare it with the auxiliary connection module. Based on the comparison results, obtain the combination with the highest fit of the secondary sorting results in the auxiliary connection module to obtain the target combination.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This trusted data element circulation system based on blockchain and artificial intelligence is implemented collaboratively by a demand retrieval module, a feature ranking module, and a data preheating module. This effectively reduces the burden of on-chain operations and real-time query response time. The demand analysis module adopts a federated learning framework and zero-knowledge proofs, generating demand vectors and exchanging gradient parameters only off-chain through intelligent agents, reducing the need for direct on-chain processing of large amounts of data. Furthermore, the demand retrieval module sets retrieval thresholds to break down large-scale queries into single searches, and combines these with the elimination thresholds of the feature ranking module to quickly refine the target dataset. The data preheating module performs secondary sorting based on feature ranking criteria and stores the results in an auxiliary connection module. When a demander initiates a similar query again, the preheating extraction module directly compares the preheated data, eliminating the need for repeated full-process retrieval and reducing system resource consumption.

[0015] Meanwhile, the demand analysis module supports multi-demand federated learning, extracting shared demand features by exchanging gradient parameters, thus avoiding bias from single demands. The feature ranking module splits and ranks the target dataset based on demand feature attributes, and filters low-match results by removing thresholds to ensure that the returned data is highly relevant. At the same time, the secondary sorting of the data preheating module creates independent storage pages, and the preheating extraction module directly maps the target combination in subsequent queries to achieve "preheated retrieval," which greatly shortens the waiting time and improves system availability. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 This is a schematic diagram of the target dataset acquisition process of the present invention; Figure 3 This is a schematic diagram of the target dataset acquisition structure of the present invention; Figure 4 This is a schematic diagram of the sorting information item acquisition process of the present invention; Figure 5 This is a schematic diagram of the sorting results of the sorting information items in this invention. Detailed Implementation

[0017] 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.

[0018] This application provides a trusted data element circulation system based on blockchain and artificial intelligence. Through the collaborative implementation of a demand retrieval module, a feature ranking module, and a data preheating module, it effectively reduces the burden of on-chain operations and real-time query response time. The demand analysis module employs a federated learning framework and zero-knowledge proofs, generating demand vectors and exchanging gradient parameters only off-chain via intelligent agents, reducing the need for direct on-chain processing of large datasets. Furthermore, the demand retrieval module sets retrieval thresholds, breaking down large-scale queries into single searches, and, combined with the feature ranking module's elimination thresholds, quickly refines the target dataset. In the implementation case, for a demand of 1000 data points, the system filters invalid results using thresholds, shortening retrieval time. The data preheating module performs secondary ranking based on feature ranking criteria and stores the results in an auxiliary connection module. When a demander initiates a similar query again, the preheating extraction module directly compares the preheated data, eliminating the need for repeated full-process retrieval and reducing system resource consumption. Figure 1 As shown, it includes a requirements analysis module, a requirements retrieval module, a feature sorting module, a data preheating module, and a preheating extraction module.

[0019] Requirements Analysis Module: Obtains data from the demand side, performs feature analysis and feature extraction to obtain requirement feature items.

[0020] It is important to note that the demand feature term is used to represent the data demand of the demand party based on the trusted space. The methods for obtaining the demand feature term include: based on the trusted space, with blockchain as the trust technology foundation, the demand party generates a demand vector through an off-chain smart agent and uploads the commitment value verified by zero-knowledge proof to the blockchain to obtain the uploaded information term; the trusted space is based on artificial intelligence, and through the data platform party and the demand party jointly execute a federated learning framework, based on the uploaded information term, the gradient parameters are exchanged and features are extracted in the local model training to obtain the demand feature term.

[0021] It is important to note that when a demander needs to obtain data, since data in the trusted data circulation system is transmitted within a trusted space, it is necessary to perform feature analysis and feature extraction on the demander's data requirements, and then perform data association within the trusted space based on the extracted features.

[0022] In the specific implementation process, a community health service center in a certain city needs to obtain the "core characteristics of diabetes risk" from residents' health data, such as blood glucose fluctuation patterns and medication adherence. At this time, the demand side generates a demand vector through an off-chain smart agent and uploads the commitment value verified by zero-knowledge proof to the blockchain to obtain the uploaded information item. The process is as follows: sequenceDiagram Demand-side intelligent agent -> Local AI model: Input demand description text Note the right of the local AI model: "Identifying features such as abnormal blood sugar levels and medication intervals". Local AI model -->> Demand-side intelligent agent: Generates a 128-dimensional demand vector [0.87, -0.23, ...] Demand-side intelligent agent ->> ZKP generator: Creates vector commitment values ZKP Generator -> Blockchain: Upload Commitment = Hash(Vector + Random Salt) Blockchain -->> Smart Contracts: Generating Requirement NFTs #301 Trusted Space utilizes an artificial intelligence approach, where a data platform collaborates with demanders to execute a federated learning framework. The federated learning architecture is shown in Table 1. Table 1 Role Execute action Blockchain Interaction Hospital A 1. Retrieve the required NFT#301 metadata. 2. Initialize the feature extraction model. Access to payment amount Hospital B / C 1. Load resident health data locally. 2. Calculate model gradients ▽WB, ▽WC. Submit gradient hashes to on-chain storage Aggregation Contract Perform secure aggregation: ▽Wagg = Σ(▽Wi × weight i) Triggering a verifiable random function (VRF) to select nodes NFT#301 is the unique identifier for the on-chain requirement feature task, ▽W is the partial derivative of the loss function with respect to the weights, B and C are the local gradient vectors of node B / C, which are used to carry local data feature information but do not expose the original data record, and ▽Wagg is the global aggregate gradient, which integrates knowledge from multiple nodes and is the key input for updating the shared model parameters. Based on the uploaded information items, gradient parameters are exchanged and features are extracted during local model training to obtain the required feature items. Taking the PySyft framework as an example, during the feature extraction stage: # Hospital B Local Implementation model=LinearRegression() local_data=load_encrypted("hospital_B_data.enc") # Only exchange gradients (original data does not leave the local machine) grads=model.compute_gradients(local_data,target_vector=NFT#301) submit_to_chain( data_hash=sha256(grads), zk_proof=prove_correct_computation(grads) ) Obtain the output results and generate and validate the required feature items: { "Feature ID":"FT-20250829-301", "Core Features": [ {"name":fasting blood glucose fluctuation coefficient","weight":0.92,"source": "HospitalA"}, {"name":"Frequency of Missed Dosage of Hypoglycemic Medications","weight":0.87,"source": "HospitalB"}, {"name":"Monthly BMI Change Slope","weight":0.76,"source": "HospitalC"} ], "Verify evidence": [ "Zero-knowledge proof verification passed: zkp_7a3d19" "Gradient Consistent Hash: 0x8f3e...cda2" ] }

[0023] It should be noted that the method for obtaining demand feature terms also includes: dynamically selecting participating nodes through a verifiable random function to obtain participating node terms; obtaining the demand vector of participating node terms and uploading the commitment value verified by zero-knowledge proof to the blockchain to obtain uploaded information participating terms; executing a federated learning framework based on uploaded information participating terms; obtaining at least two local domains based on the local information of uploaded information participating terms; and obtaining demand feature terms by exchanging gradient parameters and performing feature extraction in local model training based on artificial intelligence through local domains.

[0024] It is important to note that when there is more than one demander, it is necessary to dynamically select the demander as a participating node, obtain the demand vector of each participating node, and use the same steps to exchange gradient parameters and perform feature extraction in the local model training to obtain demand feature terms. This achieves the common data retrieval and feature extraction effect for each demander.

[0025] In the specific implementation process, the transnational banking consortium needs to identify the characteristic patterns of fraudulent trade finance. Because data is scattered across heterogeneous nodes such as ports, customs, and logistics companies, and there is competition between these nodes, data collusion must be prevented. Furthermore, compliance requirements vary significantly across different regions. Therefore, VRF (Virtual Ranking System) is used to dynamically select participating nodes. graph TB A [The requester initiates a feature extraction task] --> B (calls the VRF smart contract) B-->C { Input seed parameters:} Block height #1290382 Task ID: T-20250829 C-->D [Output random number RN=0x7a3d...f29e] D-->E [Sorted by node staked token weight] E-->F [Select the cluster with node RN mod N] F-->G [Announcement of participating nodes:] • Port of Rotterdam (EU) • Singapore Customs (ASEAN) •Maersk Logistics (Company) Then, the gradient parameters are exchanged and features are extracted again during local model training to obtain the required features.

[0026] Demand retrieval module: Performs feature retrieval based on demand features to obtain the target dataset.

[0027] It is important to note that, such as Figure 2 As shown, the target dataset includes at least one retrieved data. The target dataset is used to represent the associated data in the trust space that matches the features of the demand feature item. The method for obtaining the target dataset includes: obtaining the feature type information of the demand feature item and the feature demand information corresponding to the feature type information; performing a single search based on the single feature type information and the corresponding feature demand information; setting a search threshold, which is a fixed search difference between the single feature type information and the corresponding feature demand information, with the search threshold being 20%; obtaining the result set of the single search; and thus obtaining the target dataset.

[0028] In the specific implementation process, such as Figure 3As shown, the current data requirement of the demander is to obtain the "core features of diabetes risk" from 50,000 sets of resident health data within one year. At this time, the demander needs to obtain the feature type information of the required feature item and the corresponding feature requirement information. The feature type information is "core features of diabetes risk", and the corresponding feature requirement information is "50,000 sets" and "within one year". According to the set search threshold, the feature requirement information is 50,000-60,000 sets and 0-292 days. The target dataset is obtained by using "core features of diabetes risk" 50,000-60,000 sets and "core features of diabetes risk" 0-292 days as a single search.

[0029] Feature ranking module: Based on the demand feature items, the target dataset is ranked to obtain ranking information items. The search data with the first position in the ranking result is used as the benchmark demand data of the demand feature items to obtain the demand feature data that matches the demand feature items.

[0030] It is important to note that, such as Figure 4 As shown, the method for obtaining ranking information items includes: performing feature classification based on demand feature items to obtain at least one demand feature category information; obtaining the demand restriction threshold corresponding to the demand feature category information, including quantity information and time information, to obtain demand feature attribute information corresponding to the demand feature category item; splitting the target dataset based on the demand feature attribute information; obtaining data information in the target dataset that conforms to the demand feature attribute information based on the splitting result; using the conformity of the data information as a criterion to perform feature ranking to obtain ranking result information; setting a removal threshold, which is a fixed percentage value, the removal threshold is 10%; removing the last information in the ranking result information by a percentage based on the removal threshold to obtain the remaining ranking information, thus obtaining the ranking information item.

[0031] In the specific implementation process, such as Figure 5 As shown, the current data requirement of the client is to obtain the "core features of diabetes risk" from 1000 sets of residents' health data within one year. The required feature type information and corresponding feature requirement information are obtained. The feature type information is "core features of diabetes risk," and the corresponding feature requirement information is "1000 sets" and "within one year." Based on the set search threshold, the required feature information is 1000-1200 sets and 0-292 days. Using "core features of diabetes risk" (1000-1200 sets) and "core features of diabetes risk" (0-292 days) as single searches, ten hospitals provided data in the target dataset. The results are shown in Table 2. Table 2 Table of contents Feature Requirements Information 1 Feature requirement information 2 Hospital 1 1100 0-365 Hospital 2 1100 0-365 Hospital 3 1200 0-365 Hospital 4 1100 0-365 Hospital 5 1300 0-280 Hospital 6 980 0-270 Hospital 7 900 0-270 Hospital 8 1100 0-250 At this point, the matching items of the data information are used as the criterion for feature sorting, and the sorting result information is: Hospital 8 > Hospital 5 > Hospital 6 > Hospital 7 > Hospital 3 = Hospital 4 = Hospital 2 = Hospital 1. Based on the set elimination threshold of 10%, Hospital 1 is eliminated to obtain the sorting information item.

[0032] Data preheating module: Based on the feature ranking criteria used to sort the target dataset, the retrieved data in the auxiliary connection module is sorted and stored to obtain stored preheating data.

[0033] It is important to note that the secondary sorting is based on the weight percentage data of the feature sorting criteria used in the feature sorting. The features sorting criteria are used as the main weights for sorting, and the results of the secondary sorting are stored to obtain the pre-warmed storage data, thereby realizing the pre-warming storage of the data combination.

[0034] It should be noted that the secondary sorting method includes: obtaining the demand feature data corresponding to the first search data in the sorting information item as the main judgment feature; obtaining at least one feature to be judged based on the feature splitting result of the target dataset; and sorting the target dataset by using the feature to be judged as the main judgment feature in turn to obtain at least one secondary sorting result.

[0035] In the specific implementation process, the current data requirement of the client is to obtain the "core characteristics of diabetes risk" from 1000 sets of resident health data within one year. At this time, the client needs to obtain the feature type information and the corresponding feature requirement information. The feature type information is "core characteristics of diabetes risk," and the corresponding feature requirement information is "1000 sets" and "within one year." Based on the set search threshold, the required feature information is 1000-1200 sets and 0-292 days, respectively, using "core characteristics of diabetes risk" (1000-1200 sets) and "diabetes risk" as examples. Using the "core features" (0-292 days) as a single retrieval, the target dataset was obtained from three hospitals: Hospital A provided 1000 sets of "core features of diabetes risk" within 366 days, Hospital B provided 2000 sets within 280 days, and Hospital C provided 800 sets within 360 days. The initial ranking was Hospital A > Hospital B > Hospital C. However, the primary determining feature was the number of sets, and the secondary determining feature was time. Using the secondary determining feature as the primary determining feature, the target dataset was ranked, resulting in a secondary ranking: Hospital B > Hospital A > Hospital C.

[0036] It should be noted that the method for obtaining the preheating data includes: based on the combination results of the target dataset, the trusted space creates an auxiliary connection module, which includes at least one auxiliary storage page; based on the secondary sorting results, the sorting results are independently split according to the main weight category, resulting in at least two secondary sorting items, which are stored in the auxiliary storage pages respectively, with the main weight category as the auxiliary label of the auxiliary storage page.

[0037] Preheating and extraction module: Once the data requirements are obtained again, feature analysis and feature extraction are performed again, and the results are compared with those of the auxiliary connection module to obtain the target combination.

[0038] It is important to note that feature analysis and feature extraction are performed again to obtain the updated feature items. Based on the updated feature items, the target combination is obtained by comparing them with the auxiliary connection module. Based on the target combination, the data is mapped to the target auxiliary storage page and extracted and presented, thereby realizing the pre-warming retrieval and extraction of data.

[0039] It should be noted that the method for obtaining the target combination includes: based on the updated feature items, obtaining the updated feature type information and the updated feature requirement information corresponding to the updated feature type information, obtaining the update requirement limit threshold of the updated feature requirement information, including quantity information and time information, comparing it with the auxiliary connection module, and obtaining the combination with the highest fit of the secondary sorting result in the auxiliary connection module according to the comparison result, thus obtaining the target combination.

[0040] It is important to note that since the secondary sorting results stored in the auxiliary connection module are sorted based on the feature sorting criteria as the sovereign weight and stored independently, each feature in the feature sorting has a highest-ranking independent storage result. At this time, based on the update requirement limit threshold of the updated feature requirement information and the comparison result of the auxiliary connection module, the sorting result can be directly obtained from the stored auxiliary storage page, thereby achieving the effect of data preheating storage and rapid extraction of data combinations.

[0041] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. A trusted data element circulation system based on blockchain and artificial intelligence, comprising: Demand Analysis Module: Acquires data requirements from the demand side, performs feature analysis and feature extraction to obtain demand feature items, which are used to represent the demand side's data requirements based on the trust space; Its characteristic is that it further includes: Demand retrieval module: Based on demand feature items, feature retrieval is performed to retrieve at least one data item and obtain the target dataset. The target dataset is used to represent the associated data in the trust space that matches the features of the demand feature items. Feature ranking module: Based on the demand feature items, the target dataset is ranked to obtain ranking information items. The search data with the first position in the ranking result is used as the benchmark demand data of the demand feature items to obtain the demand feature data that matches the demand feature items. Data preheating module: Based on the feature ranking criteria used to rank the target dataset according to the required feature items, the retrieved data in the auxiliary connection module is sorted a second time. The basis for the second sorting is the weight ratio of the feature ranking criteria used for feature ranking. The data is sorted according to the feature ranking criteria as the main weight, and the results of the second sorting are stored to obtain the stored preheated data, thereby realizing the data preheating and storage of the data combination. Preheating extraction module: The data requirements are obtained again, and feature analysis and feature extraction are performed again to obtain updated feature items. Based on the updated feature items, the module compares them with the auxiliary connection module to obtain the target combination. Based on the target combination, the module maps to the target auxiliary storage page and performs data extraction and presentation, thereby realizing the preheating retrieval and extraction of data.

2. The trusted data element circulation system based on blockchain and artificial intelligence according to claim 1, characterized in that: The methods for obtaining the demand feature items include: Based on the Trust Space and with blockchain as the foundation of trust technology, the demand side generates a demand vector through an off-chain smart agent and uploads the commitment value verified by zero-knowledge proof to the blockchain to obtain the uploaded information item. Trusted Space uses artificial intelligence to implement a federated learning framework in conjunction with the data platform and the demand side. Based on the uploaded information items, gradient parameters are exchanged and features are extracted in the local model training to obtain the required feature items.

3. A trusted data element circulation system based on blockchain and artificial intelligence according to claim 2, characterized in that: The method for obtaining the demand feature items also includes: By dynamically selecting participating nodes through a verifiable random function, participating node items are obtained, the demand vector of participating node items is acquired, and the commitment value verified by zero-knowledge proof is uploaded to the blockchain to obtain the uploaded information participating item. The federated learning framework is executed based on the uploaded information participants. Based on the local information of the uploaded information participants, at least two local domains are obtained. Based on artificial intelligence, gradient parameters are exchanged and features are extracted in the local model training through the local domains to obtain the required feature terms.

4. A trusted data element circulation system based on blockchain and artificial intelligence according to claim 1, characterized in that: The methods for obtaining the target dataset include: Obtain feature type information of the required feature items and feature requirement information corresponding to the feature type information, and perform a single search based on the individual feature type information and the corresponding feature requirement information; Set a retrieval threshold, which is a fixed retrieval difference between the single feature type information and the corresponding feature requirement information, obtain the result set of a single retrieval, and then obtain the target dataset.

5. A trusted data element circulation system based on blockchain and artificial intelligence according to claim 1, characterized in that: The method for obtaining the sorting information item includes: Based on the demand feature items, feature classification is performed to obtain at least one demand feature type information. The demand restriction threshold corresponding to the demand feature type information is obtained, including quantity information and time information, and the demand feature attribute information corresponding to the demand feature type item is obtained. The target dataset is split based on the demand feature attribute information. Based on the splitting result, the data information in the target dataset that meets the demand feature attribute information is obtained. The matching items of the data information are used as the criterion to sort the features and obtain the sorting result information. Set a rejection threshold, which is a fixed percentage value. Based on the rejection threshold, the last item in the sorting result information is removed by a percentage to obtain the remaining sorting information and thus obtain the sorting information item.

6. A trusted data element circulation system based on blockchain and artificial intelligence according to claim 1, characterized in that: The secondary sorting method includes: The requirement feature data corresponding to the first search data in the sorting information item is obtained as the main judgment feature. Based on the feature splitting results of the target dataset, at least one feature to be judged is obtained. The target dataset is sorted by using the features to be judged as the main judgment features in turn, and at least one secondary sorting result is obtained.

7. A trusted data element circulation system based on blockchain and artificial intelligence according to claim 1, characterized in that: The method for obtaining the stored preheating data includes: Based on the combined results of the target dataset, the trusted space creates an auxiliary connection module, which includes at least one auxiliary storage page; Based on the secondary sorting result, the sorting result is independently split according to the main weight category, resulting in at least two secondary sorting items. The secondary sorting items are stored in the secondary storage pages respectively, and the main weight category is used as the secondary tag of the secondary storage page.

8. A trusted data element circulation system based on blockchain and artificial intelligence according to any one of claims 1-7, characterized in that: The method for obtaining the target combination includes: Based on the updated feature items, obtain the updated feature type information and the updated feature requirement information corresponding to the updated feature type information. Obtain the update requirement limit threshold of the updated feature requirement information, including quantity information and time information. Compare it with the auxiliary connection module. Based on the comparison results, obtain the combination with the highest fit of the secondary sorting results in the auxiliary connection module to obtain the target combination.

Images